Us and Them: The Role of Group Identity in Explaining Cultural Resonance and Framing Effects

ABSTRACT

This study investigates the importance of group identity as a determinant of cultural resonance in news framing effects research. Two parallel survey experiments were conducted in the United States and Britain; respondents received a news story about a military atrocity attributed to either American or British soldiers during the war in Afghanistan. In one condition, the story contained only basic facts about the incident. In the other, the story included frames reinforcing the national identity of the accused nation, including: (a) downplaying the incident; (b) denigrating the transgressors; and (c) redirecting attention toward positive aspects of the nation. We find that these frames tend to be more powerful when protecting ingroup (versus outgroup) identity. The reason, we argue, is due to their cultural resonance, not simply the factual information that they make salient. More broadly, our results offer important insights into the framing process, illuminating the psychological mechanisms that help explain how cultural resonance can facilitate specific responses among audiences, particularly in response to group identity threats.

The influence of national identity on news reporting has been well-documented. Reporters, despite professional norms and routines that favor objectivity and neutrality (Schudson, 1978; Tuchman, 1978), have been found, time and again, to embrace news narratives that serve to bolster the nation, including its military and citizens, in a variety of contexts (e.g., Fishman & Marvin, 2003; Gans, 1979). During times of war, for example, these dynamics have been shown to be particularly evident, from news coverage of isolated military actions (Entman, 1991) to ongoing military conflicts (Slattery & Doremus, 2012; Tirman, 2011; Wolfsfeld et al., 2008). Most notably, ethnocentrism in news coverage tends to be particularly pronounced when military atrocities occur—those incidents in which members of the military deliberately and systematically harm civilians (e.g., Finlay, 2018; Rowling et al., 2011). Studies have shown, for example, that in such moments, journalists often uncritically amplify the frames put forth by officials, which tend to focus on appealing to and protecting the national identity. This tends to occur even when dissent is vigorously expressed by other credible official sources (e.g., Rowling et al., 2015).

While these studies have sought to define the parameters of when and why ethnocentric news coverage tends to manifest in response to military atrocities, much less attention has been devoted to the broader effects of this news coverage on public opinion. To the extent that scholars have explored these dynamics, such studies have tended to rely on public opinion polls or, in some cases, experiments to demonstrate how these frames—e.g., downplaying the incident, denigrating the transgressors and positively restoring the national identity—affect attitudes about military incidents and the nation more broadly (e.g., Castano, 2008; Rowling et al., 2015). Evidence that these frames are effective in these situations can also be derived from the fact that national discourse in the aftermath of military atrocities—e.g., the My Lai massacre (1968), the Abu Ghraib prison scandal (2004) and the Kandahar massacre (2012), to name a few—tends to rapidly return to a focus on the underlying rationale behind the war and the virtuousness of the nation without a broader and more thorough examination of the causes and consequences of these incidents (Bennett et al., 2008; Jones & Sheets, 2009; Rowling et al., 2011).

Nonetheless, the underlying psychological mechanisms that might render citizens receptive to these frames remain unclear. Indeed, there appears to be a theoretical puzzle within this broader area of research that remains unaddressed: Do these frames “work” among citizens in these situations primarily because of their logical coherence and plausibility (e.g., such incidents are exceedingly rare, a few “bad apples” were responsible, such incidents are not reflective of the principles and values of the nation) or rather because they appeal to and serve to protect the broader national identity? Put another way, do citizens embrace these frames because the considerations they make accessible to audiences (e.g., Cacciatore et al., 2016; Price & Tewksbury, 1997) are convincing based on their perceived factual merits and the sources involved, or is it something more deeply rooted in social psychology—that these frames resonate with audiences’ values and identities (e.g., Baden & David, 2018; Gamson, 1992; Gamson & Modigliani, 1989), encouraging citizens to feel better about their nation and, by extension, themselves in response to these atrocities?

To explore these dynamics, we conducted two parallel 2 × 2 factorial survey experiments with samples of American and British citizens to assess the manner and extent to which these frames shape how citizens respond to military atrocities. Specifically, respondents from both countries were exposed to a news article about an incident—derived from real events—in which a small group of either American or British soldiers killed and desecrated the bodies of Afghan civilians during the war in Afghanistan. The news article was either presented in a short, factual newswire style, or included frames that downplayed the incident, denigrated the transgressors and positively restored the national image. This design enabled us to assess whether the effects of these frames were universal across all respondents or more impactful among those citizens whose military and, by extension, national identity was implicated in the transgressions.

Overall, we find that national identity is indeed a crucial factor in how respondents react to news stories about military atrocities. The effects of the frames were more pronounced when respondents were exposed to a news story in which their military committed the transgressions (versus the other national military). We also found that respondents’ embrace of the ingroup frames was mediated by the extent to which they felt a personal sense of embarrassment/shame in response to the news story. These results shed valuable light on why military atrocities so often go unpunished and, in some cases, even unnoticed among those citizens whose national identity is implicated in such incidents. More broadly, these findings offer important insights into the manner and extent to which national identity serves as an important determinant in whether news frames become culturally resonant and, therefore, significantly shape how audiences perceive and respond to a specific event. In this way, our study contributes in meaningful ways to our understanding of the underlying psychological mechanisms that lead to framing effects. Group identity is central to these dynamics, we argue, especially in moments when the image and reputation of the nation is perceived to be threatened. We clarify and elaborate on these concepts and ideas below, and we reflect on the implications of these findings for journalists and citizens.

National identity and the news

Studies have widely demonstrated the profound impact of national identity—a sense of belonging and attachment to the nation—on news reporting. Daily reinforcement of the nation through “banal nationalism” (Billig, 1995), domestication processes (e.g., Clausen, 2008), and the basic guiding news values of proximity and relevance (Harcup & O’Neill, 2017) ensure that foreign news is reliably conveyed to domestic audiences in ways that prioritize national actors and themes. Such coverage reinforces citizens’ national identity and gives credence to the notion that nations are the natural political ordering of the world (Billig, 1995; Gellner, 1983). These patterns, studies suggest, become most pronounced and potentially controversial in times of international conflict. In such moments, more blatantly skewed, ethnocentric reporting of military incidents and transgressions tends to occur (e.g., Bennett et al., 2008; Entman, 1991; Finlay, 2018; Ginosar & Cohen, 2017; Wolfsfeld et al., 2008). In particular, political and military officials are often given priority as news sources in these moments—due to their access, expertise and credibility on such matters (Zaller & Chiu, 1996)—and, thus, receive a platform to manage the political narrative related to these incidents. As previous scholarship has suggested, such deference to and relatively unfiltered transmission of the frames put forth by these officials in such moments is due, in large part, to economic (Hutcheson et al., 2004), cultural (Gans, 1979), and psychological (Baden & David, 2018; Wolfsfeld et al., 2008) influences on the news production process. Put another way, the nature of news production leads journalists to engage, often unintentionally, in ethnocentric news coverage of issues or events, especially those that are perceived to threaten the image and reputation of the nation.

These dynamics, we argue, can be particularly problematic within a democratic society, given that citizens must rely upon the news media to inform them about international events, to offer critical perspectives on government policies and interventions, and to hold those in power accountable for their actions at home and abroad (Althaus et al., 1996)—especially when those actions are morally questionable. Numerous studies, however, have demonstrated that these ideals are rarely achieved in news coverage during times of conflict. Instead, as reflected in Bennett’s (1990) “indexing hypothesis,” Entman’s (2004) “cascading activation model,” Wolfsfeld et al.’s (2008) notion of “defensive reporting,” and Ginosar and Cohen’s (2017) concept of “patriotic journalism” (see also Bennett et al., 2008; Hutcheson et al., 2004; Zaller & Chiu, 1996), what we often see in these situations is a heavy reliance on, and deference to, government sources. As a result, the public is exposed to a largely uncritical promotion of official interpretations of issues and events within mainstream news. While any normative critique of the press, including whether it adheres to the notion of objectivity, is necessarily culturally situated (Donsbach & Klett, 1993), such uncritical coverage of the nation in times of conflict, especially when military atrocities occur, raises important moral and ethical questions about the role of the press within a democracy (see Bucy & D’Angelo, 1999).

News frames, cultural resonance, and social identity

Numerous studies have employed the concept of framing to examine the manner and extent to which officials and journalists seek to emphasize and convey—or push—their interpretations of issues or events into the public arena. A critical axiom in the framing literature is that not all frames are created equal. For example, Entman et al. (2009) emphasize that frames, by definition, must connect to existing culture—or the “stock of schemas commonly found in the minds of a society’s individuals” to make sense of information and the world around them (p. 176). But certain frames are understood to be more culturally resonant than others (Baden & David, 2018; Entman, 2004; Gamson, 1992; Gamson & Modigliani, 1989). This is based on whether the frame effectively activates or elicits receptive thoughts and emotions among its audience. As Gamson (1992) argues:

[S]ome frames have a natural advantage because their ideas and language resonate with a broader political culture. Resonances increase the appeal of a frame by making it appear natural and familiar. Those who respond to the larger cultural theme will find it easier to respond to a frame with the same sonorities. (p. 135)

Indeed, frames that tap into and engage with prevailing cultural beliefs and values will stand the best chance to influence individual perceptions and achieve broader success within the public arena, “cascading” (see Entman, 2004) past any resistance or opposition that might arise from other political actors or within the press (Chong, 2000; Druckman & Nelson, 2003). As Entman (2004) argues: “The most inherently powerful frames are those fully congruent with schemas habitually used by most members of society” (p. 14).

Recently, scholars have nuanced the discussion of resonance further, recognizing that it has both a factual, knowledge-based component, as well as a normative, value-based one (Baden & David, 2018; Simonsen, 2019). Whereas the former focuses on the alignment between a frame and the prior beliefs or cognitions of its audience, the latter pertains to the alignment between a frame and the values and attitudes of its audience (in Baden and David’s (2018) words, a frame’s “epistemic plausibility” versus its “normative appropriateness” (p. 530)). When frames match audience perceptions along both dimensions, they tend to “click” with the audience and become easily accepted (Baden & David, 2018, p. 514). When frames diverge along one or both dimensions, it leads to varying degrees of adaptation, debate or outright rejection. In particular, when frames tap into the normative-cultural dimension but do not reaffirm established values or identities, a vigorous debate then tends to ensue (Baden & David, 2018). And any attempt to legitimize controversial actions through such non-resonant frames is less likely to succeed (Simonsen, 2019).

Our study aims to further specify the psychological mechanisms driving the normative dimension of resonance—that is, this alignment between a frame and the relevant identities at stake. In particular, we suggest that social identity is the key to understanding this normative resonance. As scholarship on social identity suggests, citizens tend to derive comfort, self-esteem and security from their identification with particular social groups (Tajfel, 1982; Tajfel & Turner, 1986). This often leads individuals to engage in favoritism toward their own social group (Brewer, 1999; Dasgupta, 2004) and prejudice toward other social groups (Sherif, 1966; Smith & Bond, 1999). As Triandis (1994) has argued, it is “natural to help or cooperate with members of our in-group, to favor our in-group, to feel proud of our in-group, and to be distrustful of and even hostile towards out-groups” (pp. 251–252). Notably, this ingroup bias has been shown to be particularly pronounced when the ingroup is perceived to be threatened (Branscombe & Miron, 2004; Entman, 1991). In such scenarios, members of the threatened social group tend to behave in ways that seek to protect or enhance the group identity (Tajfel & Turner, 1986; Wohl & Branscombe, 2008, 2009). Indeed, given that the social group serves as a source of comfort, self-esteem and security, its maintenance and preservation is crucial for its members.

National identity, in particular, is recognized as one of the most potent forms of group identity because it “commands profound emotional legitimacy” among citizens (Anderson, 1983/2006, p. 4). It is through the lens of national identity, for example, that citizens tend to process and interpret global issues and events. In the context of military atrocities, however, when the image and reputation of the nation are perceived to be threatened, such incidents, we argue, are likely to elicit among citizens a profound sense of uncertainty and collective angst (see Wohl & Branscombe, 2008, 2009). In such contexts, citizens are forced to reconcile their beliefs in the principles and virtues of the nation with the potential shame and humiliation likely caused by these transgressions. As such, the types of frames, we argue, that are most likely to normatively resonate within the citizenry in such moments are those that serve to protect and reinforce the national identity.

Research in social psychology (see Bandura, 1990, 1999), for example, has shown common responses among ingroup members when threats to group identity arise from within the ingroup: minimization of the deviance; dehumanization of the victims; moral justification of ingroup actions; denigration of the perpetrators; and advantageous comparisons (see also Branscombe & Miron, 2004; Castano, 2008; Marques & Paez, 1994). As Bandura (1990, p. 37) has argued: “As long as the detrimental results of one’s conduct are ignored, minimized, distorted, or disbelieved, there is little reason for self-censure to be activated.” Extant scholarship has also shown that three specific group identity-protective frames tend to manifest in political and news discourse when these types of incidents occur—minimization, disassociation and reaffirmation (see Entman, 1991; Finlay, 2018; Rowling et al., 2011, 2015; Tirman, 2011; Wolfsfeld et al., 2008).

First, minimization involves characterizing ingroup deviance as anomalous and/or blaming the behavior on relatively insignificant group members, which serves to downplay the scope and severity of the transgression (Bandura, 1999; Blatz et al., 2009; Grey & Martin, 2008). This allows the group to limit or outright avoid aversive emotional reaction and guilt triggered by the behavior. Second, disassociation consists of taking measures to remove or punish the deviant ingroup actors by, for example, characterizing them as “un-American” or vowing to punish them (Eidelman et al., 2006; Marques & Paez, 1994). In effect, the process of casting out deviant actors allows the ingroup to preserve its positive group identity (Baden & David, 2018). Third, reaffirmation shifts attention away from the transgression to aspects of the group that portray it in a much more positive manner (Bandura, 1990; Tajfel, 1982). Often this involves an emphasis on idealized group values, attributes and behavior (Billig, 1995). The result, inevitably, is that members of the group reestablish themselves in a preeminent position relative to other groups, which serves to restore positive social identity.

Most notably, these specific frames have been observed in content analyses of U.S. political and news discourse following the 1968 My Lai Massacre (Rowling et al., 2015), the 2004 Abu Ghraib prison scandal (Rowling et al., 2011), and in the discourse within Israel in response to the 2014 Gaza War (Finlay, 2018). These frames have also been shown to manifest within the discourse when foreign civilians have been killed in specific military strikes (see Entman, 1991; Wolfsfeld et al., 2008) and in the discourse surrounding broader war efforts (Tirman, 2011). Psychologically and discursively, then, it seems there is plenty of evidence for the prevalence and potential appeal of such frames in times when the identity of the nation is perceived to be threatened.

Because such frames derive from basic social identity motivations and occur robustly in news and political discourse, it is important to understand precisely how they work, and for whom. The framing techniques of minimizing the severity of the incidents, casting the perpetrators out of the group, and reaffirming the positive group image are highly likely to “resonate” with ingroup members, we argue, because they are not only factually plausible to their audience, but they also offer normative reinforcement of the moral values and identity of the nation. Thus, these frames should resonate both epistemologically and normatively (Baden & David, 2018; Simonsen, 2019). To test this proposition—that such frames are prevalent and powerful in moments of national dissonance because of this normative dimension—we need to parse the factual, epistemological plausibility of such frames from their normative resonance. This is precisely what our study does—comparing responses to such frames among ingroup and outgroup audiences. In the case of military atrocities, we expect that those respondents whose national identity has been implicated in the transgressions (i.e. the ingroup conditions) would be more receptive to the frames because of their alignment with the social psychological needs and values of the group; in contrast, those receiving these frames about an outgroup have no such national identity at stake and, therefore, we argue, are less likely to embrace the frames, however logically coherent and plausible they might seem.

In effect, our experiment allows a test of this dual resonance framework. If the power of these frames lies in their epistemic, factual assertions—and, thus, their strength were not dependent upon their resonance with respondents’ group identities—we should expect essentially no difference in how respondents react to the frames between the ingroup and outgroup conditions. Conversely, if the power of these frames lies in the normative dimension—that is, the extent to which they tap into respondents’ positive sense of group identity—we should see much stronger effects in the ingroup versus the outgroup conditions—evidence that the frames are culturally resonant and “click” more with the ingroup audience than an outgroup one (Baden & David, 2018).

Before we specify our hypotheses, we feel it useful to briefly explain the experimental set-up and relevant attitudes. We conducted two simultaneous experiments: one in the United States and the other in Great Britain. Respondents in both countries were given a news story about an alleged American or British military atrocity during the war in Afghanistan. Each news story contained either basic facts about the incident (control condition)¹ or the same description of the incident accompanied by frames designed to bolster the national identity of the implicated nation (framed condition). We then focused on the extent to which respondents: (a) expressed personal embarrassment/shame; (b) sought to disassociate—or cast out of the ingroup—the soldiers involved in the incident; and (c) actively reaffirmed positive aspects of the nation involved. These measures were designed to assess how the incident affected respondents’ own personal sense of (positive) group identity and whether the frames were adopted; more details about these items are provided in the methods section.

Our first three hypotheses focus on the control condition, assessing the impact of whether the transgressions (without any accompanying frames) were committed by ingroup versus outgroup soldiers. This was done to establish a baseline for ingroup-versus-outgroup comparisons, against which the framing effects could be examined (see Chong & Druckman, 2007). Because the ingroup control message includes negative information about the ingroup, respondents should perceive a clear threat to their identity and, in response, take the psychological steps necessary to restore that identity. We expect this to manifest in three ways. First, ingroup respondents should experience a greater sense of embarrassment/shame compared to those receiving the outgroup control message (see Gross, 2008). Second, respondents should express more negative attitudes toward the ingroup transgressors compared to those in the outgroup control message (as this allows such transgressors to be “kicked out” of the ingroup, psychologically; see Bandura, 1990, 1999; Marques & Paez, 1994). Third, respondents who receive the ingroup control message should, nevertheless, reaffirm their group identity by exhibiting greater positivity toward their ingroup than toward the outgroup in the other condition. Thus, these hypotheses (and variables) are primarily based on basic ingroup/outgroup identity dynamics. The ingroup control message (versus the outgroup control message) should lead to a stronger sense of negative personal involvement in the incident, but also a greater protection of the ingroup. Concisely, our initial set of hypotheses are: Respondents who receive the ingroup (versus outgroup) control message will be more likely to express embarrassment/shame (H1a), disassociate the perpetrators (H1b), and positively reaffirm the national identity (H1c).

Next, we examine how the frames affect ingroup versus outgroup respondents. These frames, derived directly from previous research, actively downplay the severity of the incident, denigrate the transgressors, and positively restore the national image (Bandura, 1990, 1999; Castano, 2008). Because of this, we expect the frames to lead to a decrease in personal embarrassment/shame for ingroup respondents (H2a), an increase in disassociation of the perpetrators (H2b) and a boost in positive reaffirmation of the national identity (H2c), relative to the control message. Notably, these effects, we argue, should be moderated by ingroup/outgroup membership. That is, because the national-identity-protective frames should resonate both factually and normatively with ingroup respondents, their effects should be stronger for the ingroup than the outgroup (H3). Thus, we predict stronger framing effects in the ingroup (versus outgroup) conditions.

As a final theoretical step, we attempt to empirically demonstrate a mechanism explaining these proposed normative resonance effects. In facing allegations of ingroup transgressions, respondents should experience an identity crisis of sorts: evidence that the ingroup is bad when the fundamental ingroup image is (generally) good. We expect this conflict to manifest in the embarrassment/shame measure, which should mediate the ability of respondents to positively reaffirm the national identity—again, as moderated by the ingroup/outgroup variable. We select the reaffirmation dependent variable as the focal one here, because it represents a broader measure of positive group identification than the more specific disassociation variable (see also the explanation of the variables in the methods section). Indeed, several studies (see Gross, 2008; Gross & D’Ambrosio, 2004; Kim & Cameron, 2011) have shown how frames can elicit among respondents powerful and varied emotional reactions, which tend to be conditioned by both predispositions (e.g., national identity) as well as the content of the frame. These emotional responses can then, in turn, significantly impact information processing and opinion formation among respondents. We therefore expect that the effect of the frames on respondents’ willingness to positively reaffirm the national identity will be mediated by their embarrassment/shame, but only for ingroup members (H4).

Method

Design and stimuli

We conducted two parallel 2 × 2 factorial survey experiments. Respondents (who were either American or British, and who first gave informed consent before entering the survey) were asked to read a news story, derived from real events and attributed to the Associated Press, about a military incident involving soldiers from either the United States or Britain accused of killing innocent civilians in Afghanistan, then desecrating their bodies (Boal, 2011; Leigh & Evans, 2010). The news article was either presented in a short, factual newswire style (control condition), or accompanied with frames designed to downplay the incident, denigrate the transgressors and positively restore the national image (framed condition). Thus, respondents received a news story about soldiers from either their ingroup, or an outgroup, committing military atrocities. The full text of each of these manipulations are available in the digital-only supplemental file.²

Procedures

After reading the news story, respondents answered questions gauging their emotional responses to the message, including whether they felt any embarrassment and/or shame in response to the news story. These items were followed by measures tapping into incident-related attitudes (e.g., how isolated was the incident, who should be blamed, etc.). Respondents were then asked for demographic information, followed by questions about their level of national attachment, trust in the government and military of the country in question, and how believable they found the message. The full text of these questions is available in the supplemental file; mean scores and scale reliabilities are reported later in this section.

Sample

After the study received ethical approval, respondents were recruited through SurveyMonkey, an international company that administers surveys to targeted populations.³ Those who participated in the survey had the opportunity to enter a “sweepstakes” for a chance to earn monetary credits from SurveyMonkey to be given to charities of their choice. We did not provide any additional incentives for taking the survey. Because respondents self-selected into our study, they are not representative of the American or British adult populations. Nonetheless, respondents were recruited by SurveyMonkey to be comparable in age, sex, and geographical region to these populations, and more diverse than traditional student samples. Demographic data for each sample is in the supplemental file. We ran the experiment simultaneously in the United States and Britain on August 9–15, 2016, via two different web platforms of Survey Monkey. The datasets were then merged. On average, the survey took 11 minutes to complete and 69% of respondents finished the survey.

In total, 1,676 respondents completed the experiment. Given the aversive nature of our manipulation, we assessed the extent to which respondents found the message believable, on a 6-point scale from “not at all” to “completely.” Eighty-three respondents (5.0%) answered “not at all” on this scale, indicating total rejection of the message. After careful consideration, we decided to filter these respondents out of subsequent analyses, leaving us 1593 respondents in total.⁴ The distribution (N) by conditions was as follows: N_{US control} = 391, N_{US frames} = 391, N_{UK control} = 412, N_{UK frames} = 399. Overall, 794 respondents (49.8%) were American; 799 were British. If a respondent received a message about soldiers from his or her own national military, it was coded as an “ingroup” message. In total, 791 respondents (49.7%) received an ingroup message; 802 received an outgroup message. Randomization based on major demographic variables (age, sex, income, political ideology, religion) was successful.

Dependent variables

There are three dependent variables: “Embarrassment/Shame,” “Disassociation,” and “Reaffirmation.” Embarrassment/Shame is an index based on two, 6-point Likert-type items (ranging from 1 “not at all” to 6 “very much so”): the extent to which the respondent felt “embarrassed” or “ashamed” while reading the stimulus. These items were standardized onto 0–1 scales and then averaged (α = .926, M = .55, SD = .34). Disassociation was also based on two 6-point items, which derive from the language in the manipulations themselves and which are based on the literature outlined in the theoretical section (Eidelman et al., 2006; Marques & Paez, 1994): the extent to which respondents found that “the soldiers involved in this incident have brought shame to” the country in question, and the extent to which “the soldiers involved in this incident deserve to wear” the military uniform of their country (reverse coded). These items were standardized on 0–1 scales and averaged (α = .732, M = .77, SD = .26). Finally, reaffirmation comprised seven 6-point items, indicating respondents’ opinions about the nation, its military, and its foreign policy. These measures were derived from literature on ingroup protective strategies (Bandura, 1990; Tajfel, 1982), and from previous work on the presence of such frames in news discourse (Bennett et al., 2008; Entman, 1991; Rowling et al., 2011; Wolfsfeld et al., 2008). They included items such as “Do you think that America/Britain adheres to international law in its conduct during war?” and “When it comes to solving world problems, the United States/Britain usually makes things worse” (reverse-coded). The items were standardized on 0–1 scales and averaged (α = .91; M = .59, SD = .23).

Results

Our first set of hypotheses (H1a–c) expected that, when faced with an unframed control message about a military atrocity committed by the ingroup (versus the outgroup), respondents will be more likely to express embarrassment/shame, to disassociate the perpetrators and positively reaffirm the national identity. Independent-sample t-test results are shown in Table 1.

Table 1. Comparing ingroup and outgroup control messages on incident-related attitudes.

	Ingroup Control (N = 406)	Outgroup Control (N = 397)	T-Test (Ind. Samples, 2-tailed)	Cohen’s d
Embarrassment/Shame	M = .63 (SD = .34)	M = .52 (SD = .35)	t = 4.39, df = 801, p < .001	d = .32
Disassociation	M = .72 (SD = .29)	M = .78 (SD = .25)	t = 2.78, df = 791, p < .01	d = .22
Reaffirmation	M = .60 (SD = .22)	M = .52 (SD = .23)	t = 5.04, df = 801, p < .001	d = .36

The results in Table 1 show support for hypotheses H1a and H1c; an ingroup control message leads to higher embarrassment/shame among respondents as well as more positive reaffirmation of the national identity compared to the outgroup control message. We did not, however, find support for H1b; in fact, respondents were significantly more likely to disassociate outgroup transgressors compared to ingroup transgressors for identical transgressions. We return to this finding in the subsequent analyses.

Next, we compared the control and framed messages, hypothesizing that the frames would lead to a decrease in personal embarrassment/shame (H2a), an increase in disassociation of the perpetrators (H2b) and a boost in positive reaffirmation of the national identity (H2c), relative to the control message—especially among ingroup members (H3). Essentially, we expect these frames to factually and normatively resonate with the ingroup—allowing them to “click” more strongly for the ingroup than the outgroup. For ease of interpretation, we present first the independent-samples t-test results for ingroup members in Table 2.

Table 2. Comparing framed to control conditions among ingroup respondents.

	Ingroup Control (N = 406)	Ingroup Framed (N = 385)	T-Test (Ind. Samples, 2-tailed)	Cohen’s d
Embarrassment/Shame	M = .63 (SD = .34)	M = .57 (SD = .33)	t = 2.2, df = 789, p < .05	d = .18
Disassociation	M = .72 (SD = .29)	M = .78 (SD = .24)	t = 2.84, df = 778, p < .01	d = .23
Reaffirmation	M = .60 (SD = .22)	M = .65 (SD = .21)	t = 2.61, df = 789, p < .01	d = .23

Here, H2a–c are supported. Respondents receiving the frames express less embarrassment/shame than their control condition counterparts, and they are more likely to disassociate the perpetrators from the national group, and to positively reaffirm the national identity. Table 3 presents the same results for the outgroup.

Table 3. Comparing framed to control conditions for outgroup respondents.

	Outgroup Control (N = 397)	Outgroup Framed (N = 405)	T-Test (Ind. Samples, 2-tailed)	Cohen’s d
Embarrassment/Shame	M = .52 (SD = .35)	M = .49 (SD = .35)	t = 1.24, df = 800, p = .216	d = .09
Disassociation	M = .78 (SD = .25)	M = .80 (SD = .24)	t = 1.70, df = 800, p = .089	d = .08
Reaffirmation	M = .52 (SD = .23)	M = .57 (SD = .24)	t = 3.09, df = 800, p < .01	d = .21

H3 expected that these findings would be stronger for the ingroup than outgroup conditions. Indeed, looking at Table 3, we see that the frames still have an effect in the outgroup conditions, though it is weaker, on respondents’ disassociation (marginally significant) and reaffirmation measures (significant). Specifically, respondents are more likely to denigrate the outgroup transgressors and positively reaffirm the national identity of the other nation when exposed to the frames in the outgroup conditions. Interestingly, the personal, emotion-related impact of the frames is absent in the outgroup conditions: Receiving outgroup frames has no effect on the sense of embarrassment or shame of outgroup respondents. We also ran 2-way ANOVAs for each dependent variable, to examine whether ingroup identity statistically moderated the effect of the frames on these attitudes. These analyses, however, did not reach statistical significance.⁵ Therefore, we can only give our resonance-focused H3 partial support. That said, it is important to note that there is one consistent finding across our analyses: Although we only ever see partial and mixed effects on outgroup members, we see consistent and significant effects on those in the ingroup. We expand on this below.

Finally, we examine the relationship between personal embarrassment/shame and positive reaffirmation of national identity to further parse the resonance mechanism. We expected in H4 that embarrassment/shame would mediate the effects of the frames on the extent to which respondents positively reaffirm the national identity, but only for the ingroup conditions. This is consistent with the idea that these frames normatively align with preexisting positive group identity among ingroup members. Given that we see framing effects on reaffirmation in both ingroup and outgroup conditions, we viewed this as, in effect, a test of whether normative or factual resonance better explains the framing effects here. To explore these dynamics, we used Hayes (2018) PROCESS macro to examine the direct and indirect effects of the frames on reaffirmation. We first ran model 4 (10,000 bootstrap samples) for the basic mediation, separated by ingroup and outgroup, for ease of interpretation. The results are presented in Figure 1.

Figure 1. Mediation analyses, for ingroup and outgroup conditions.

Note. Bold numbers indicate significant paths.

As the top half of Figure 1 shows, we find a significant indirect effect of the frames, mediated through Embarrassment/Shame, for the ingroup messages, b = 0.006, 95% CI [0.001, 0.012]. The confidence interval of the standardized indirect effect (b = .013, 95% CI [.002, .028]) does not include zero; thus, significance can be concluded. In contrast, the lower half of Figure 1 shows that only the direct effect of the frames on reaffirmation is significant for outgroup messages. Specifically, this effect is not mediated by Embarrassment/Shame, and the total model is not significant. The indirect effect has a 95% CI [–0.004, 0.006] that includes zero and is thus not significant.

We then confirmed these results via a moderated mediation model (model 59, again 10,000 bootstrap samples), which tests the mediation model as moderated by ingroup/outgroup on all 3 paths. We see that for the outgroup, the indirect (mediated) model is insignificant (b = .0014, 95% Bca CI [−0.0004, 0.0059], whereas the ingroup indirect model shows a significant mediation (b = .0057, 95% Bca CI [.0009, 0.0124]. This suggests that for the ingroup message, the frames lead to less embarrassment/shame, and lower embarrassment/shame leads to higher reaffirmation (or more embarrassment/shame leads to less reaffirmation). It should be noted, however, that the index of moderated mediation is not significant in this analysis (b = .0043, 95% Bca CI [−.0016, .0115]). Nonetheless, as Hayes (2015) himself points out, such conflicting findings are difficult to interpret and do not definitively undercut the notion that these findings are significant. Thus, we found conditional support for H4.

Discussion

This study explores the role of group identity within the framing process, focusing on its importance in explaining how and why certain frames are culturally resonant in moments of national dissonance—specifically, military atrocities. In doing so, we engage the broader scholarly debate over framing effects and attempt to illuminate the psychological mechanisms that help explain how cultural resonance can facilitate specific responses among audiences in these situations. Overall, we find that frames designed to bolster the national identity of those implicated in such transgressions tend to prompt particular reactions among national ingroup (versus outgroup) members. In particular, such frames allow ingroup members to alleviate a personal discomfort that they feel when confronted with allegations of ingroup transgressions; the frames also encourage respondents to (further) disassociate the alleged transgressors from the ingroup, and to reaffirm a positive group identity—classic responses to group deviance (Bandura, 1990, 1999). That the role of personal embarrassment/shame in this response was particularly important for ingroup-but not outgroup-respondents, suggests that these frames align (or resonate) with a preexisting positive group identity among ingroup audiences, going farther than simply offering a factually plausible account of what happened (Baden & David, 2018; Simonsen, 2019). Indeed, the psychological comfort these messages seem to provide to ingroup audiences, in particular, lends credence to the notion that group identity was central to why and how these frames resonated.

A few limitations are worth noting here. First, this was an experiment offering constructed news stories about a hypothetical incident (though based on real incidents). The amount of information given to respondents and the artificiality of the account likely impacted our findings. This may account for the relatively small effect sizes. But the connection between our constructed stimuli and pervasive news patterns observed in previous work makes the external validity of our study higher than it might otherwise appear. Second, our sample, though large, is not representative of the national population and, as always, an online experiment lacks control where it gains in some external validity relative to a lab study; our struggle with the 83 respondents who found the message “not at all believable” illustrates this problem (see footnote 4). Future studies should build in more careful attention and robustness checks to avoid risks of conditioning on posttreatment variables (Montgomery et al., 2018). Third, our stimuli had different lengths (see footnote 2); while we do not believe it impacted our results, future studies should take this into consideration. It is certainly possible that the amount of time spent immersed in the message within the framing conditions affected responses independently of the content itself; encouragingly, however, the ingroup-outgroup differences are primarily examined within the same type of message—either control or framed—meaning that length was consistent between comparison groups. Nonetheless, future research should attempt to construct externally valid stimuli with more similar lengths, and above all, should include better attention checks and pacing (see Malhotra, 2008) when possible. Fourth, we acknowledge that the U.S. and U.K. are culturally similar contexts (Inglehart et al., 1996); this may have made the outgroup messages more (personally) affecting even for outgroup respondents, and therefore made the outgroup framing effects still present, if weaker, on several measures. Indeed, we suspect this may have played an important role in undercutting the hypothesized moderation effects. Future research examining outgroup comparisons with less culturally similar, less favored or even actively disliked outgroups might help elucidate these patterns and further define the parameters of what types of frames are culturally resonant. Finally, perhaps the largest threat to external validity for our study is in the single-exposure nature of this message; when such incidents occur, they are widely publicized (albeit typically briefly), and respondents are likely inundated with messages reaffirming group identity in response; repeated-stimulus studies would be welcome.

Future research should build upon this idea of group identity-based cultural resonance. The psychological mechanism, we argue, pertains to cognitive dissonance: when confronted with allegations of ingroup transgressions, respondents feel some level of discomfort and attitudinal imbalance (Harmon-Jones & Mills, 2019). We suggest that our embarrassment/shame measure taps into that cognitive dissonance. Our analyses suggest that only in the ingroup conditions was embarrassment/shame meaningfully attenuated by the frames; that attenuation, in turn, allowed respondents to engage in more positive reaffirmation of the national group. Thus, these frames assuage personal identity discomfort which allows restoration of a positive group image and balance with a positive self-concept. Future work examining and measuring such dynamics with groups of varying positivity for respondents could help illuminate whether this dissonance mechanism applies.

If a group-identity-based understanding of culturally resonant frames is viable, however, it can be studied in other contexts as well. Think of any group identity-based political context—e.g., racial violence, inter-faith or inter-ethnic conflict. When members of one group are accused of harming those in another, we would expect, discursively, the same group identity protective frames to emerge and succeed in messages designed for that (first) group. Our design could be directly replicated in such contexts, even if the complexities of intersectionality (e.g., Bowleg, 2008) and multi-level group identities would need further attention. Overall, we believe this study contributes not only to the scholarly debate on framing effects and mechanisms, but also to societal understanding of just how such identity-protective frames seem to arise and succeed in response to national transgressions. It is precisely in these moments, we argue, that we need critical and open debate over policies; instead, we often receive reassuring messages about the fundamental goodness of our nation, allowing citizens to move on without any meaningful change.

Acknowledgments

The authors would like to thank the University of Nebraska-Kearney for providing a collaborative research grant that facilitated this research.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by the University of Nebraska-Kearney [Collaborative Research Grant].

Notes on contributors

Penelope Sheets

Penelope Sheets is a senior lecturer in the Department of Communication Science at the University of Amsterdam. Her research and teaching focus on political communication, (data) journalism, and public opinion, with a particular interest in the role national, racial, religious and other group identities play in the construction and flow of news, as well as how group identity affects citizens’ responses to political communication.

Charles M. Rowling

Charles M. Rowling is a professor of political science at the University of Nebraska at Kearney. His research focuses on the role of political communication, news coverage, and public opinion in U.S. foreign policy. Much of his work has examined these dynamics in the context of national identity.

Jason Gilmore

Jason Gilmore is an associate professor of global communication at Utah State University. His research examines the strategic creation, dissemination, and effects of particularly potent national and international ideas—such as American exceptionalism, patriotism, anti-Americanism—that are regularly communicated within and across national borders.

Niklas Melcher

Niklas Melcher completed his research master degree at the University of Amsterdam with a focus on political communication, globalization and intercultural connections. Nowadays, he works in market research in the cultural industries.

Notes

¹ It is debatable whether a truly “unframed” message can exist; every story is told from a certain angle, and every set of facts presents a certain lens on an issue. However, we refer to the control condition as “unframed” relative to the framed condition—the latter involves distinct identity-protective frames, whereas the control condition merely recounts the alleged events without any explanation or crisis management.

² Despite the differences in length of stimulus—which, to a certain extent, are unavoidable given the complexity of designing a plausible control condition that neither reaffirms the national identity at stake in the message, nor provides more factual details that would fundamentally change the manipulation—analyses show that respondents did not spend significantly more time participating in the experiment in the framed versus control conditions (though completion time by itself should not be taken as an indication of survey performance (Malhotra, 2008), nor are these data sufficiently complete to conduct a proper analysis for our datasets). Perhaps more importantly, we did not see any difference in the priming of respondents’ national attachment, which was an essential component of the theoretical mechanisms at play. Reported national attachment was statistically identical between the framed and control conditions (5.20 (control, SD = 1.7) versus 5.25 (framed, SD = 1.6); t = .599, df = 1585.6, p = .549. Notably, this is a fairly high reported mean score on the measure, which suggests that both messages were capable of priming national attachment despite varying lengths.

³ IRB application number: 031815-4, University of Nebraska at Kearney, approval granted March 19, 2015.

⁴ Respondents were most likely to reject the message in the British control condition (N = 33). We ran all analyses with these additional 83 respondents included. Analyses for H1a-c are nearly identical. Analyses for H2/H3, however, differ: the effects of the ingroup frames on Embarrassment/Shame and on Reaffirmation are no longer significant. Similarly, the mediation results are also insignificant with these message rejecters included. We debated at length how to interpret these findings and proceed. While a self-reported total rejection of the message (as “not at all believable”) could, at face value, be taken as a form of identity protection—that is, one rejects a message about one’s group (indeed, 24 of the 33 rejecters in the UK control condition were UK citizens) because one finds it so unpalatable—the findings for this particular subgroup do not suggest this to be the case. If the rejection were a form of group protection, we should find even stronger effects of the frames, which alleviate that group identity discomfort, among this subset of respondents. This is not the case. Furthermore, we ran the analyses with the bottom two response categories, and the bottom three response categories (all below the mid-point of the scale) excluded. In this case, the results remain similar to the exclusion of just the bottom category—in some cases, the findings are slightly stronger, but always directionally identical. Given the similarity among these categories of message “doubters,” and their distinct difference from those true message “rejecters” (who answered “not at all believable”), we chose to continue with only the lowest group dropped, because that answer option does seem conceptually different to us than to have rated a non-labeled 2 or 3 on that 6-point scale. For transparency, side-by-side analysis comparisons with these 83 respondents are available in the digital-only supplemental file. Analyses with the full sample, and the various sub-samples based on these exclusions, are also available from the corresponding author.

⁵ EmbShame: Overall F = 12.63, df(3,1), p < .001; interaction term F(ingroup*framed) = 0.382, p = .537; Disassociation: Overall F = 10.57, df(3,1), p < .001; interaction term F(ingroup*framed) = 0.901, p = .341; Reaffirmation: Overall F = 19.92, df(3,1), p < .001; interaction term F(ingroup*framed) = 0.227, p = .634.