1. Introduction
Well, I think that when you get right down to it, we’re a nation that speaks English. I think that, while we’re in this nation, we should be speaking English.
Trump, 2015 (THR Staff 2015)
The U.S. Census Bureau (2015) projections suggest that by the mid-21st century, the U.S. will be heavily populated by racial minorities, in which non-Hispanic Whites will constitute less than 50% of the entire U.S. population. In their story, CNN (2008) emphasized that “by 2050, 54% of the population will be minorities.” In particular, the Asian American and Hispanic populations are expected to increase from 15.5 million to 40.6 million and from 46.7 million to 132.8 million, respectively (CNN 2008). Both the Asian American and Hispanic populations are projected to nearly triple from 2012 to 2060 (U.S. Census Bureau 2012). Exposure to this U.S. demographic shift makes White people more politically conservative, support tougher immigration policy, endorse diversity less, and express more anger and fear toward minorities (Craig and Richeson 2014; Danbold and Huo 2015; Outten et al. 2012, 2018). Putnam (2007) shows that individuals living in areas that are high in diversity and densely populated with immigrants tend to “hunker down”; that is, displaying lower trust in local government, altruism, and perceived quality of life. In the short run, increased immigration and diversity results in decreased social solidarity or a growing sense of fragmentation (Putnam 2007). The ongoing demographic shift in which Whites are becoming another minority is perceived by White Americans as a threat to their group’s social status (Craig and Richeson 2014; Outten et al. 2012, 2018) and their prototypicality as Americans (Danbold and Huo 2015) which accounts for the effects.
The changing U.S. demographics are reflected in the increasing presence of non-English speaking and bilingual immigrants and has put language at the center of political debates regarding multiculturalism and immigration (Newman et al. 2012). Indeed, a large portion of Latinx and Asian immigrants hold onto their native language; 62% and 36% of Latinx Americans and Asian Americans, respectively, speak no English or mostly some other language at home (Sears and Savalei 2006).1 The presence of non-English speakers in the U.S. may challenge a perceived fundamental aspect of what it means to be American (e.g., Citrin et al. 1990; Citrin and Wright 2009; Schildkraut 2007), leading many Americans to experience “culture shock” in their own land (Oberg 1960).
1.1. America = White (and Speaking English)
An underrated component of research on demographic shifts is the construction of American identity such that these “shifts” are linked to cultural changes that are threatening. Several lines of research suggest that what it means to be American is equal to what it means to be White (Devos and Banaji 2005; Devos and Mohamed 2014; Sidanius et al. 1997; Staerklé et al. 2010). Devos and Banaji (2005) found that Asian American and Black American exemplars were both regarded as less American than White American exemplars on an implicit attitude test (IAT). In other research, Sidanius and colleagues (1997) have shown that U.S. White respondents have higher levels of patriotism and a higher positive association between racial and national identity compared to U.S. racial minorities. This subgroup asymmetry points to a greater sense of “ownership” over what it means to be American for Whites compared to Blacks as a function of the former racial subgroup’s status and power. A greater sense of ownership over the nation affords certain representations of history that glorify the nation (Kurtis et al. 2010; Liu and Hilton 2005) and institutionalizing cultural values such as “independence” that are consonant with White, middle-class ways of being (Markus 2017).
Another construction of U.S. identity is linked to “being able to speak English”—a marker of cultural assimilation to Anglocentric definitions (see Citrin et al. 2001; Mukherjee et al. 2013, 2018; Pehrson and Green 2010). Citrin et al. (2001) have shown that most Americans believe that the ability to speak English is important for being “truly American.” Specifically, 93 percent of the 1996 General Social Survey of the American public aged 18 and older deemed it “very” or “fairly” important for a “true American.” In fact, foreign-born U.S. residents whose native language is not English are as likely as White and Black Americans—whose native tongue is often English—to consider the ability to speak English as a criterion for American identity (Citrin et al. 1990).
1.2. Exposure to a Foreign Language as a Threat
Research on the constructions of American identity and immigration suggests that “speaking English” is an implicit but critical component of what it means to be American (see Gerteis et al. 2020). Given this assumption, the present research asks what happens when individuals are exposed to a foreign language in a U.S. context? There is a small but growing literature examining this question, and it employs the theoretical framework of group-based threats (Riek et al. 2006; Stephan et al. 2015) to illuminate how exposure to a foreign language results in White Americans expressing antipathy toward immigrants. These frameworks include realistic threat, symbolic threat, assimilationist threat theory, and social identity threat (Branscombe et al. 1999; Paxton and Mughan 2006).
1.2.1. Realistic Threat
Realistic threat refers to the potential harm to the physical, material, or economic well-being of one’s ingroup (Mukherjee et al. 2013; Stephan et al. 1999). It often stems from perceived competition over scarce economic resources and perpetuates the notion that immigrants “steal” jobs from dominant group members, resulting in anti-immigrant sentiments (Esses et al. 2001; Stephan et al. 1999). Realistic threat is included in the present research as it is a major perspective within immigration research and, to a lesser degree, on foreign language exposure (e.g., Newman et al. 2012).
1.2.2. Symbolic Threat
Symbolic threat pertains to violations to a group’s core set of values, morals, beliefs, standards, and cultural norms (Kinder and Sears 1981; Stephan et al. 1999). It is a threat to one’s ingroup’s worldview, often caused by their ethnocentric belief that their ingroup is superior to outgroups. Those who do not share the ingroup’s worldview are seen as threatening, and this threat results in hostility toward them. As such, Americans may feel threatened by immigrants and hold negative attitudes toward them if they are not seen as assimilating to American norms (e.g., speaking English). We include this measure in the present research given the symbolic (i.e., cultural) threat that White individuals may experience upon hearing a language other than English in a public space in the U.S.
1.2.3. Assimilationist Threat Theory
Assimilationist threat theory posits that the failure of immigrants to successfully adapt to American culture (e.g., ability to speak English) is culturally threatening. It argues that a vast majority of Americans believe that immigrants should be able to communicate effectively in English and that they should only speak in English when in the presence of other Americans. When immigrants are not seen to be doing so, English-speaking Americans feel frustrated, excluded, and culturally threatened (see Newman et al. 2012). Paxton and Mughan (2006) reported that participants strongly felt that immigrants were being rude and even arrogant for conversing in their foreign language in public places in the presence of English-speaking Americans. In another study, non-Spanish speaking Americans who imagined being excluded from a conversation in Spanish at a workplace reported low organizational commitment, experienced symbolic threat, and displayed prejudice toward immigrants (Hitlan et al. 2006). In follow-up research, Hitlan et al. (2010) found that non-Spanish speaking Americans who were excluded from a conversation in Spanish at a non-workplace setting felt angry and expressed prejudice toward immigrants. These lines of research tap into the epistemic concerns of cultural threats by exposing participants to a foreign language they do not understand. However, none of the studies have directly measured epistemic threats in the context of exposure to a foreign language.2 The present research fills this gap in the literature. Furthermore, the emphasis on epistemic threat is what distinguishes this perspective from symbolic threat.
1.2.4. Social Identity Threat
Social identity threat concerns threats to the value of one’s social identity (Branscombe et al. 1999). Past research has suggested that social identity threats to one’s ingroup can prompt individuals to defend themselves by expressing intergroup bias against outgroups (e.g., Branscombe and Wann 1994; Cadinu and Reggiori 2002; Leach et al. 2003). For instance, Branscombe and Wann (1994) exposed American participants to one of two short videos of a boxing match—where one version induced a threat to the value of being an American by portraying the American boxer as losing to his Russian opponent—and found that exposure to social identity threat resulted in outgroup derogation. Given that being American is often conceptualized as being White and being able to speak English, White Americans may feel that their value of being a White American is threatened when exposed to a foreign language and express anti-immigrant attitudes as a result. To test this, we create and utilize a measure specific to the vignette itself to assess participants’ social identity threat.
Although social identity threat—namely threat to the value of one’s social identity—and symbolic threat may be similar (e.g., Riek et al. 2006; Stephan et al. 2015), their underlying mechanisms may be different. Social identity threat stems from wanting to maintain a positive group image/esteem—primarily focusing on the ingroup—whereas symbolic threat is more concerned with the nature of threats posed by the outgroup (Rios et al. 2018). In the case of exposure to a foreign language, White Americans could experience social identity threat as immigrants speaking in their native tongue could come off as immigrants devaluing and disrespecting “American” values—for example, being White and speaking English. White Americans could simultaneously experience symbolic threat, where the use of a foreign language comes off as “changing” what America means. No research to our knowledge has examined social identity threat in tandem with symbolic threat; in addition, we examine these and other threats (e.g., epistemic threat) for the first time in the literature on exposure to foreign language.
1.3. Contributions of the Present Research
Past studies on exposure to a foreign language have largely focused on exposure to Spanish and the implications of how White Americans respond to it and perceive immigrants. What has largely been ignored is whether exposure to different foreign languages affects White Americans’ feelings of threat and attitudes toward immigrants differently. The current research fills this gap by experimentally manipulating the type of foreign language participants are exposed to—Korean (Studies 1 and 2), Spanish (Study 1), or German (Studies 1 and 2)—in order to examine the differential effect that being exposed to a different kind of foreign language has on Whites’ experiences of group-based threats and attitudes toward immigrants. Moreover, in Study 2, we manipulate the salience of epistemic threat with the inclusion (vs. exclusion) of subtitles to see whether making the conversation understandable dampens threat—particularly, epistemic threat—the participants experience from being exposed to a foreign language. The focus here is on exposure to a foreign language in a U.S. context given the ongoing, rapid growth of the Asian American and Hispanic populations in the U.S. The U.S. is one of the most linguistically homogeneous countries, where a vast majority of individuals only know English (Newman et al. 2012), which may in part explain why the ability to speak English is a crucial component of American identity to begin with and controversies surrounding the use of languages other than English remain more persistent than other Western countries also faced with impending racial/ethnic demographic shifts.
In line with previous research, we argue that exposure to a foreign language leads White Americans to form less positive perceptions of the targets and their conversation, experience group-based threats, and heightens their anti-immigrant attitudes. This is because immigrants speaking in their native tongue would appear as violating one of the key constructions of being American (i.e., speaking in English) and as culturally threatening the American way of life. Moreover, we argue that exposure to either Korean or Spanish would lead them to form less positive perceptions of the targets and their conversation and greater group-based threats and anti-immigrant attitudes than exposure to German, as the former two do not map onto an Anglocentric construction of American identity, whereas the latter is in line with racialized constructions of American identity (e.g., being White).
Spanish vs. Korean
One might argue that exposure to Spanish would be more threatening and thus lead to greater anti-immigrant attitudes because the sheer number of Latinx immigrants outweigh that of Asian immigrants (CNN 2008; U.S. Census Bureau 2012). The U.S. immigration discourse has focused on undocumented immigrants from Mexico and Central American countries and has propagated the notion that they are a problem with regards to criminality, violence, and depletion of resources (Chavez 2008; De Genova 2005). Alternatively, it might be that exposure to Korean would be more threatening and lead to greater anti-immigrant attitudes because Asian Americans have long been seen as perpetual foreigners who will never be able to fully assimilate into the mainstream American culture (Armenta et al. 2013). Thus, the current research examined whether there is a difference in outcomes between the two foreign languages that are both inconsistent with Anglocentric constructions of American identity.
1.4. Overview of Present Studies
In two experiments, we examined if exposure to a foreign language would lead White Americans to form less positive perceptions of the targets and their conversation, experience group-based threats, and heighten their anti-immigrant attitudes. In Study 1, White Americans were randomly assigned to read one of four fictional transcripts of a conversation of an immigrant family at a restaurant. Depending on the condition they were assigned to, the type of language being spoken by the family was manipulated to be either Korean, Spanish, German, or English. In Study 2, White Americans read the same fictional transcript—minus the Spanish; however, there was an addition of two subtitles conditions in which the subtitles were provided next to the Korean and German texts. After reading one of four (Study 1) or five (Study 2) vignettes, participants completed measures pertaining to the conversation (e.g., anti-immigrant attitudes). Based on past research on group-based threats, in Study 1 we predicted that exposure to a foreign (i.e., Korean, Spanish, or German) vs. the English language would lead to decreased positive perceptions of the targets and their conversation and increased group-based threats and anti-immigrant attitudes. Moreover, we predicted that White Americans’ feelings of threat (i.e., social identity threat) would account for the relationship between foreign language exposure and increased anti-immigrant sentiment. Likewise, in Study 2 we predicted that exposure to a foreign (i.e., Korean or German—with or with no subtitles) vs. English language would lead to decreased positive perceptions of the targets and their conversation and increased group-based threats and anti-immigrant attitudes. We predicted that White Americans’ feelings of threat would mediate this relationship as in Study 1. Furthermore, we predicted that exposure to a foreign language with subtitles and English would decrease epistemic threat. We also predicted that there would be a difference within different types of foreign languages, in that exposure to either Korean or Spanish would lead to lower positive perceptions of the targets and their conversation and greater group-based threats and anti-immigrant attitudes than exposure to German in both Studies 1 and 2.
2. Study 1
2.1. Method
2.1.1. Participants
An a priori power analysis was conducted using G*Power, which indicated that a total sample of 180 individuals would be needed to achieve a power of 0.80 with a medium effect size. One hundred seventy-three self-identified White Americans (41.0% men, Mage = 43.71 years, SD = 15.49) were recruited in early Spring 2020 via CloudResearch (Prime Panels).3 Upon completion of the study, participants received compensation in the amount they have agreed to with the platform through which they entered the survey. In line with the Declaration of Helsinki, all participants provided informed consent—by clicking “next” after viewing the consent form electronically—before beginning the study. The experiment was approved by and conducted in compliance with the University of Kansas (KU) Institutional Review Board (IRB) (STUDY00145106). Participants’ political orientation in general was moderate (M = 4.17, SD = 1.85).
2.1.2. Design and Procedure
Participants completed an online experimental study via Qualtrics in which they read a fictional transcript of a conversation between John—a recent college graduate—and his parents at a restaurant. Before being presented with the transcript, participants were asked to imagine themselves dining at Texas Roadhouse with their friends listening to country rock in the background, where they overhear part of a conversation of a nearby family. They were informed that John and his parents are U.S. immigrants and are dining to celebrate his college graduation. Throughout the conversation, John and his parents discussed a potential family trip. The type of language being spoken by John and his parents was manipulated to be either Korean, Spanish, German, or English. Participants were randomly assigned to one of the four conditions.
After reading the vignette, participants completed measures pertaining to the conversation, an attentional manipulation check, and demographic questions. Upon completion of the questionnaire, participants were debriefed and thanked.
2.1.3. Measures
All items unless otherwise stated were measured on a 7-point Likert scale from 1 (strongly disagree) to 7 (strongly agree).
Positive perceptions of the family and their conversation. Six items measured positive perceptions of the family and their conversation (e.g., “John and his parents are likeable”, “I was comfortable with the conversation of John and his parents”; α = .95).4 Higher scores reflected greater positive perceptions of the family and their conversation.
White Americans’ feelings of threat. Using three items adapted from Branscombe et al. (2007), participants rated how White Americans are likely to feel threatened in their situation (i.e., “White Americans are likely to feel distressed when listening to the conversation of John and his parents”, “White Americans are likely to feel threatened when listening to the conversation of John and his parents”, and “White Americans are likely to feel upset when listening to the conversation of John and his parents”; α = 0.96). Higher scores reflected greater perceived White Americans’ feelings of threat (i.e., social identity threat).
Anti-immigrant attitudes. To measure anti-immigrant attitudes, a feeling thermometer was used (i.e., “How do you feel toward immigrants?”). Response options ranged from 0 (very cold) to 100 (very warm). Lower scores reflected greater anti-immigrant attitudes.
Symbolic threat. Three items were adapted from the symbolic threats scale (Stephan et al. 1999) to measure threat posed by perceived differences in values and beliefs between the participants and immigrants (e.g., “Immigrants are undermining American culture”, “Immigrants should learn to conform to the rules and norms of American society as soon as possible after they arrive”; α = 0.82). Higher scores reflected greater symbolic threat.
Realistic threat. Realistic threat was measured using three items adapted from Paxton and Mughan (2006) (e.g., “Immigrants take jobs from Americans”, “Immigrants raise taxes for Americans”; α = 0.88). Higher scores reflected greater realistic threat.
Epistemic threat. Epistemic threat was measured using three items created for this study (e.g., “I feel excluded when I hear immigrants speak in their native tongue”, “I feel frustrated when I encounter immigrants with limited English”; α = 0.89). Higher scores reflected greater epistemic threat.
Manipulation check. Participants were asked to indicate whether John and his parents spoke in an Asian language (e.g., Korean, Japanese), Latin American language (e.g., Spanish, Portuguese), European language (e.g., German, French), English language, or do not recall.
2.2. Results
The means and standard deviations for Study 1’s variables are presented in Table 1. The results summarize a series of one-way ANOVAs with Tukey’s post hoc tests for each of the outcomes (also see Table 1).5
2.2.1. Positive Perceptions of the Family and Their Conversation
A one-way ANOVA for positive perceptions of the family and their conversation demonstrated that there was a significant overall effect, F(3, 169) = 27.26, p < 0.001, η2 = 0.33 (see Table 1). The follow-up Tukey post-hoc tests demonstrated that participants in the English condition (M = 6.26, SD = 0.73) perceived the family and their conversation significantly more positively than those in the Korean (M = 4.60, SD = 1.06), Spanish (M = 4.68, SD = 1.25), and German (M = 4.57, SD = 1.27) conditions. No other comparisons were statistically significant (p > 0.05).
2.2.2. White Americans’ Feelings of Threat
A one-way ANOVA for White Americans’ feelings of threat indicated there was a significant overall effect, F(3, 169) = 23.75, p < 0.001, η2 = 0.30 (see Table 1). The Tukey post-hoc tests demonstrated that participants in the English condition (M = 1.69, SD = 1.10) thought White Americans would be significantly less likely to feel threatened when listening to the conversation of John and his parents than those in the Korean (M = 3.60, SD = 1.54), Spanish (M = 3.87, SD = 1.54), and German (M = 3.93, SD = 1.76) conditions. No other comparisons were statistically significant (p > 0.05).
2.2.3. Anti-Immigrant Attitudes
A one-way ANOVA for anti-immigrant attitudes demonstrated that there was no significant overall effect, F(3, 166) = 0.08, p = 0.97, η2 = 0.002 (see Table 1).
2.2.4. Symbolic Threat
A one-way ANOVA for symbolic threat demonstrated that there was no significant overall effect, F(3, 168) = 1.50, p = 0.22, η2 = 0.03 (see Table 1).
2.2.5. Realistic Threat
A one-way ANOVA for realistic threat indicated that there was no significant overall effect, F(3, 169) = 0.95, p = 0.42, η2 = 0.02 (see Table 1).
2.2.6. Epistemic Threat
A one-way ANOVA for epistemic threat demonstrated that there was no significant overall effect, F(3, 169) = 0.02, p = 1.00, η2 < 0.001 (see Table 1).
2.2.7. Mediation by White Americans’ Feelings of Threat
Using the PROCESS macro for SPSS (Model 4; 5000 bootstrap samples; Hayes 2013), we conducted a mediation analysis where the pooled foreign language condition—the Korean, Spanish, and German conditions were combined into one foreign language condition—was compared with the English condition (dummy coded: English = 0, foreign language = 1) to examine the effects on anti-immigrant attitudes via White Americans’ feelings of threat as a mediator (see Figure 1). The foreign language condition predicted greater White Americans’ feelings of threat, β = 2.10, SE = 0.25, 95% CI [1.60, 2.60]. White Americans’ feelings of threat and the language condition were then entered simultaneously into a linear regression model predicting anti-immigrant attitudes, and the overall model was significant, F(2, 167) = 3.24, p = 0.04. Threat predicted stronger anti-immigrant attitudes after controlling for conditions, β = −3.24, SE = 1.28, 95% CI [−5.76, −0.72]. The condition did not significantly predict anti-immigrant attitudes after controlling for threat, β = 5.79, SE = 4.99, 95% CI [−4.06, 15.64]. The indirect effect of the condition on anti-immigrant attitudes via threat was significant, β = −6.80, SE = 3.12, 95% CI [−13.41, −1.25].
2.3. Brief Discussion
Study 1 presents fairly consistent evidence for measures that focus on perceptions of the targets and their conversation. In particular, exposure to a foreign (i.e., Korean, Spanish, or German) vs. English language decreases positive perceptions of the family and their conversation and increases White Americans’ feelings of threat (i.e., social identity threat). There were no statistically significant differences regarding the aforementioned outcomes between any of the foreign language conditions (all p’s > 0.05). Counter to our predictions, the participants did not make any distinctions among types of foreign language.
The findings for the measures that focus on immigrants and other group-based threats showed a lack of support for our predictions. There were no significant differences between the foreign and English language conditions on anti-immigrant attitudes, symbolic threat, realistic threat, and epistemic threat. However, exposure to a foreign language indirectly increased anti-immigrant attitudes through White Americans’ feelings of threat. Moreover, exposure to different types of foreign language did not have any differential effects on the aforementioned outcomes.
The lack of any significant effects for the latter set of outcomes may be in part attributable to the “context free” framing of the measures. Recall, we presented participants with a scenario in which they overheard part of a family’s conversation (in English or a foreign language) and asked them questions about immigrants and some group-based threats which were stripped of any mention of the family members. For instance, symbolic threat included three items assessing the threats posed by perceived differences in values and beliefs between the participants and immigrants in general. It is possible that if the items were more applicable to the “overheard conversation” (e.g., “John and his parents are undermining American culture” instead of “Immigrants are undermining American culture”), the results may be different. Thus, in Study 2, we make a subtle revision to the items regarding immigrants and group-based threats by changing the term “immigrants” to “John and his parents” when applicable.
A limitation of Study 1 is that the manipulation of English versus foreign language exposure was confounded with the participants’ ability to understand the content of the conversation between John and his parents. It is possible that the participants responded the way they did not because of the manipulation of the foreign vs. English language per se, but rather because of their (in)ability to comprehend the conversation. Thus, in addition to the original foreign language conditions, in Study 2, we added foreign language conditions with subtitles embedded to examine whether the same pattern of findings would occur.
3. Study 2
3.1. Method
3.1.1. Participants
An a priori power analysis was conducted using G*Power, which indicated that a total sample of 200 individuals would be needed to achieve a power of 0.80 with a medium effect size. Three hundred twenty-four self-identified White Americans (34.3% men, Mage = 46.56 years, SD = 16.58) were recruited in late Spring 2020 via CloudResearch (Prime Panels).6 Upon completion of the study, participants received compensation in the amount they have agreed to with the platform through which they entered the survey. In line with the Declaration of Helsinki, all participants provided informed consent—by clicking “next” after viewing the consent form electronically—before beginning the study. The experiment was approved by and conducted in compliance with the KU IRB (STUDY00145106). Participants’ political orientation in general was moderate (M = 3.89, SD = 1.84).
3.1.2. Design and Procedure
The design and procedure were similar to Study 1. As in Study 1, before being presented with the transcript, participants were asked to imagine themselves dining at Texas Roadhouse and given the same information about the family. However, there were two notable changes. First, the Spanish condition was omitted, as there was no significant difference between the Spanish and Korean conditions in all of the measures of interest. Second, in addition to the original Korean and German conditions, we added Korean and German conditions with subtitles embedded to see whether making understandable the content of the conversation would dampen any threat the participants feel from being exposed to a foreign language. The inclusion of the subtitles should moderate any epistemic threat the participants experience by allowing them to understand the conversation (i.e., with subtitles) or not understanding the conversation (i.e., without subtitles) amongst the family.
After reading the vignette, participants completed measures pertaining to the conversation, an attentional manipulation check, and demographic questions. Upon completion of the questionnaire, participants were debriefed and thanked.
3.1.3. Measures
As in Study 1, all items, unless otherwise stated, were measured on a 7-point Likert scale from 1 (strongly disagree) to 7 (strongly agree).
Positive perceptions of the family and their conversation. The same six items from Study 1 measured positive perceptions of the family and their conversation (α = 0.97).7
White Americans’ feelings of threat. Using the same three items from Study 1, participants rated how White Americans are likely to feel threatened in their situation (α = 0.96).
Anti-immigrant attitudes. To measure anti-immigrant attitudes, a feeling thermometer was used (i.e., “How do you feel toward John and his parents?”). Response options ranged from 0 (very cold) to 100 (very warm). Lower scores reflected greater anti-immigrant attitudes.
Symbolic threat. To measure symbolic threat, three items were adapted from the symbolic threats scale (Stephan et al. 1999). The items measured the threats posed by perceived differences in values and beliefs between the participants and the family (e.g., “John and his parents are undermining American culture”, “John and his parents should learn to conform to the rules and norms of American society”; α = 0.83). Higher scores reflected greater symbolic threat.
Realistic threat. Realistic threat was measured using three items adapted from Paxton and Mughan (2006) (e.g., “John and his parents take jobs from Americans”, “John and his parents raise taxes for Americans”; α = 0.88). Higher scores reflected greater realistic threat.
Epistemic threat. Epistemic threat was measured using three items created for this study (e.g., “I felt excluded by the conversation of John and his parents”, “I felt frustrated by the conversation of John and his parents”; α = 0.88). Higher scores reflected greater epistemic threat.
Manipulation checks. Participants were asked the same manipulation check question as in Study 1; however, a response option for the Spanish condition—“Latin American language (e.g., Spanish, Portuguese)”—was taken out. Participants were also asked to indicate whether the conversation between John and his parents was translated or not.
3.2. Results
The means and standard deviations for the Study 2 variables are presented in Table 2. The results summarize a series of one-way ANOVAs with Tukey’s post hoc tests for each of the outcomes (also see Table 2).8
3.2.1. Positive Perceptions of the Family and Their Conversation
A one-way ANOVA for positive perceptions of the family and their conversation demonstrated that there was a significant overall effect, F(4, 318) = 38.85, p < 0.001, η2 = 0.33 (see Table 2). The follow-up Tukey post-hoc tests indicated that participants in the Korean with subtitles (M = 6.01, SD = 1.04), German with subtitles (M = 5.76, SD = 1.27), and English (M = 5.93, SD = 1.08) conditions perceived the family and their conversation significantly more positively than those in the Korean (M = 4.38, SD = 1.32) and German (M = 4.23, SD = 0.78) conditions. No other comparisons were statistically significant (p > 0.05).
3.2.2. White Americans’ Feelings of Threat
A one-way ANOVA for White Americans’ feelings of threat indicated that there was a significant overall effect, F(4, 318) = 17.88, p < 0.001, η2 = 0.18 (see Table 2). The Tukey post-hoc tests demonstrated that participants in the Korean with subtitles (M = 2.55, SD = 1.52), German with subtitles (M = 2.93, SD = 1.55), and English (M = 1.90, SD = 1.12) conditions thought White Americans would be significantly less likely to feel threatened when listening to the conversation of John and his parents than those in the Korean condition (M = 3.78, SD = 1.52). Participants in the Korean with subtitles condition indicated significantly less White Americans’ feelings of threat than those in the German condition (M = 3.33, SD = 1.36). Participants in the German and German with subtitles conditions reported significantly greater White Americans’ feelings of threat than those in the English condition. No other comparisons were statistically significant (p > 0.05).
3.2.3. Anti-Immigrant Attitudes
A one-way ANOVA for anti-immigrant attitudes demonstrated that there was a significant overall effect, F(4, 318) = 12.58, p < 0.001, η2 = 0.14 (see Table 2). The Tukey post-hoc tests indicated that participants in the Korean (M = 63.12, SD = 27.75) and German (M = 63.54, SD = 23.65) conditions felt significantly colder toward John and his parents than those in the Korean with subtitles (M = 82.47, SD = 17.74), German with subtitles (M = 78.27, SD = 20.75), and English (M = 81.84, SD = 19.79) conditions. No other comparisons were statistically significant (p > 0.05).
3.2.4. Symbolic Threat
A one-way ANOVA for symbolic threat demonstrated that there was a significant overall effect, F(4, 317) = 12.52, p < 0.001, η2 = 0.14 (see Table 2). The Tukey post-hoc tests indicated that participants in the Korean with subtitles (M = 2.54, SD = 1.28), German with subtitles (M = 2.43, SD = 1.23), and English (M = 2.28, SD = 1.08) conditions experienced significantly less symbolic threat than those in the Korean (M = 3.41, SD = 1.56) and German (M = 3.53, SD = 1.28) conditions. No other comparisons were statistically significant (p > 0.05).
3.2.5. Realistic Threat
A one-way ANOVA for realistic threat indicated that there was a significant overall effect, F(4, 319) = 6.17, p < 0.001, η2 = 0.07 (see Table 2). The Tukey post-hoc tests demonstrated that participants in the Korean with subtitles (M = 2.18, SD = 1.25), German with subtitles (M = 2.09, SD = 1.31), and English (M = 2.09, SD = 1.17) conditions experienced significantly less realistic threat than those in the Korean condition (M = 2.94, SD = 1.59). No other comparisons were statistically significant (p > 0.05).
3.2.6. Epistemic Threat
A one-way ANOVA for epistemic threat demonstrated that there was a significant overall effect, F(4, 318) = 16.36, p < 0.001, η2 = 0.17 (see Table 2). The Tukey post-hoc tests indicated that participants in the Korean with subtitles (M = 1.98, SD = 1.12), German with subtitles (M = 1.95, SD = 1.21), and English (M = 1.72, SD = 1.10) conditions experienced significantly less epistemic threat than those in the Korean (M = 3.15, SD = 1.77) and German (M = 3.10, SD = 1.46) conditions. No other comparisons were statistically significant (p > 0.05).
3.2.7. Mediation by White Americans’ Feelings of Threat
Using the PROCESS macro for SPSS (Model 4; 5000 bootstrap samples; Hayes 2013), we conducted a mediation analysis where the pooled foreign language with no subtitles condition—the Korean and German conditions were combined into one foreign language with no subtitles condition—was compared with the pooled control condition—the Korean with subtitles, German with subtitles, and English conditions were combined into one control condition—(dummy coded: control = 0, foreign language with no subtitles = 1) to examine the effects on anti-immigrant attitudes via White Americans’ feelings of threat as a mediator (see Figure 2). As in the previous analyses, the foreign language with no subtitles condition predicted greater White Americans’ feelings of threat, β = 1.17, SE = 0.17, 95% CI [0.84, 1.50]. White Americans’ feelings of threat and the condition were then entered simultaneously into a linear regression model predicting anti-immigrant attitudes, and the overall model was significant, F(2, 319) = 46.69, p < 0.001. Threat predicted stronger anti-immigrant attitudes after controlling for the condition, β = −4.99, SE = 0.80, 95% CI [−6.57, −3.41]. The condition significantly predicted anti-immigrant attitudes after controlling for threat, β = −11.95, SE = 2.58, 95% CI [−17.03, −6.88]. The indirect effect of condition on anti-immigrant attitudes via threat was significant, β = −5.83, SE = 1.39, 95% CI [−8.94, −3.37].
3.3. Brief Discussion
In some respects, Study 2 replicates and extends Study 1’s findings. A foreign language (i.e., Korean or German—with no subtitles) not only significantly differed from English language across most of the dependent measures regarding perceptions of the targets and their conversation as in Study 1 but also from its subtitled counterparts. This suggests that participants responded the way they did in Study 1 not strictly because of the manipulation of the foreign vs. English language per se, but rather because of the unintended manipulation of the ability (vs. inability) to understand the conversation content. In other words, the inclusion of the subtitled conditions (i.e., low epistemic threat) mimicked results for the English condition for these and other dependent measures providing evidence for the epistemic threat argument regarding White Americans’ experience regarding exposure to a foreign language.
Unlike Study 1, Study 2 yielded significant condition effects on anti-immigrant attitudes, symbolic threat, realistic threat, and epistemic threat, suggesting that the lack of significant effects on outcomes emphasizing attitudes toward immigrants and the threats they pose (in Study 1) may be in part due to the framing of the items. Moreover, exposure to a foreign language with no subtitles increased White Americans’ feelings of threat, which in turn increased anti-immigrant attitudes.
4. General Discussion
With the growing presence of non-English speaking and bilingual immigrants in the U.S., the notion of “This is America, speak English” has garnered traction within the public and political discourse concerning issues of immigration and multiculturalism. A majority of the growing research on exposure to a foreign language has largely focused on exposure to Spanish and White Americans’ reactions. Moreover, research has primarily settled on outcomes such as anti-immigrant sentiment and symbolic and realistic threats. The present work corroborates the findings of past research on exposure to a foreign language (e.g., Newman et al. 2012—in which exposure to Spanish increased feelings of realistic threat and cultural threat among Americans) to some extent, but most importantly, it extends this line of research in several ways. First, it tests whether exposure to different foreign languages (Spanish, Korean, and German) affects White Americans’ feelings of threat and attitudes toward immigrants differently. Second, it examines how targets who speak in a foreign language (vs. English) are evaluated in terms of positive characteristics. Third, this is the first study that directly measures social identity threat and manipulate epistemic threat in the context of exposure to a foreign language. Social identity threat in particular has never been examined in this realm before, marking our study as the first study to incorporate it along with other major group-based threat perspectives. Furthermore, the results of the Pearson correlation indicated that social identity threat was not highly correlated with symbolic threat, r = 0.11, p > 0.05 (Study 1); r = 0.37, p < 0.01 (Study 2). This substantiates the notion that social identity threat and symbolic threat are two different constructs—or that their underlying mechanisms are different at the very least.
Across two studies, we test the impact of type of language—foreign vs. English—on White Americans’ perceptions of an immigrant family and, more broadly, attitudes toward immigrants and group-based threats. Although the condition effects on attitudes toward immigrants and some group-based threats (e.g., symbolic threat) were not consistent across Studies 1 and 2, the effects on perceptions of the targets and their conversation were consistent. Furthermore, in both Studies 1 and 2, White Americans’ feelings of threat (i.e., social identity threat) mediated the relationship between foreign language exposure and anti-immigrant sentiment. The present research fills a gap in the literature by demonstrating that different kinds of foreign language—whether they are consistent with Anglocentric constructions of American identity or not—are not perceived differently by White participants in the U.S. and that it is not the cultural foreignness of a particular language that is threatening, per se. Instead, our findings suggest individuals’ (in)ability to understand the conversation of immigrants appears to influence the perceptions they form of the targets.
4.1. Perceptions of Targets and Their Conversation
The present studies provide generally consistent support for the argument that exposure to a foreign language negatively affects White Americans’ perceptions of immigrant targets and their conversation. In Study 1, White Americans in the English condition perceived the family and their conversation significantly more positive and thought White Americans would be significantly less likely to feel threatened when listening to the conversation than those in the Korean, Spanish, and German conditions. Counter to our predictions, however, there was no significant difference among the Korean, Spanish, and German conditions. In Study 2, the findings were fairly identical in that there were significant condition effects on most of the key variables but no significant differences between the Korean and German conditions. We introduced two new conditions in Study 2, namely, the Korean with subtitles and German with subtitles conditions. We note this because there was an interesting pattern; there was no difference among the English, Korean with subtitles, and German with subtitles conditions in most of the key variables reported here. This suggests that White Americans’ (in)ability to understand the content of conversation shapes their perceptions of the targets and their conversation. As such, post hoc analyses for the perceptions of targets and their conversation in Study 2 present a more complicated story than in Study 1.
4.2. Attitudes toward Immigrants and Group-Based Threats
The present studies provide a complicated set of findings for the argument that exposure to a foreign language negatively affects attitudes toward immigrants and group-based threats among White Americans. In Study 1, there were no significant differences across the Korean, Spanish, German, and English conditions on attitudes toward immigrants or the majority of group-based threats. However, in Study 2, White Americans in the Korean and German conditions exhibited significantly greater anti-immigrant attitudes and group-based threats (e.g., symbolic threat, epistemic threat) than those in the English condition. This discrepancy in the results across the two studies is potentially attributable to the wording of the measures. In Study 1, the attitudes toward immigrants and group-based threats variables were measured using items targeting immigrants at large (e.g., “Immigrants are undermining American culture”), whereas in Study 2 these variables were assessed with revised items that were tethered to the targets in the vignette (e.g., “John and his parents are undermining American culture”). The results for these variables were significant in Study 2 but not in Study 1.
It is important to note that even after changing the wording of the items—which enabled a more sensitive test of the experimental manipulation on the measures of interest—there was no significant difference between the Korean and German conditions in Study 2. It may be that White Americans do not make nuanced distinctions between different types of foreign languages, regardless of whether it is more in line with Anglocentric constructions of American identity. Furthermore, there was no difference among the Korean with subtitles, German with subtitles, and English conditions in all of the key variables reported here, which suggests that White Americans’ (in)ability to understand the conversation of immigrants shapes their attitudes toward immigrants and experiences of group-based threats. On this note, participants experienced symbolic threat and realistic threat—in addition to epistemic threat—to the same degree in these conditions. This suggests that the manipulation of the ability (vs. inability) to understand the conversation not only prompted feelings of frustration and exclusion but also perceptions that immigrants pose threats to mainstream (White) U.S. cultural beliefs and values, economic power, and public safety. It may be the case that epistemic threat drives other group-based threats.
5. Limitations and Future Directions
The present work has several limitations. First, participants across the two studies were exposed to a foreign language via text. While exposure to a foreign language does occur in this form in real life (e.g., seeing a menu in foreign language at an ethnic food restaurant), more often it occurs through auditory channels (e.g., hearing immigrants speak in their native tongue at a restaurant). The lack of significant findings for some of the key variables and across different kinds of foreign language (e.g., Korean vs. German) may be attributed to this. That is, being exposed to a foreign language visually does not engender the same reactions as hearing a foreign language (see Srinivasan et al. 2013). Future studies should employ audio to examine the differential effect that being exposed to a different kind of foreign language might have on Whites’ experiences of group-based threats and attitudes toward immigrants. One could also try to test this by combining audio with text, where participants hear a conversation in a foreign language or English via headphones with or without English subtitles on display. This would allow one to do a full factorial design, which we were not able to do in Study 2. Study 2 only had two subtitled conditions instead of three; there was no English with subtitles condition as it would have been redundant to have the original text in English and subtitles in the same language right next to each other.
Second, it is worth acknowledging that unlike Study 1, Study 2 yielded significant condition effects on anti-immigrant attitudes and all of the group-based threats possibly due to changes in the framing of the items (i.e., “immigrants” to “John and his parents”). As such, it may be a bit of a stretch to generalize the findings concerning White Americans’ attitudes toward John and his parents and their ensuing experiences of group-based threats in Study 2 to a broader context of White Americans displaying anti-immigrant attitudes and group-based threats in general. Future studies should employ the original, general measures used in Study 1 and a modified version of the experimental paradigm that is not specific to a particular immigrant family per se, but rather about immigrants in general to assess anti-immigrant attitudes and group-based threats.
Finally, we make note that only White Americans’ feelings of threat (i.e., social identity threat) was tested as a mediator due to the temporal precedence requirement of mediation. According to the temporal precedence requirement of mediation, mediating variables must be assessed prior to the assessment of the outcome variables to demonstrate that changes in the mediating variables precede changes in the outcome variables (Kendall et al. 2017). That is, to establish mediation, the outcome variables must be measured after—not before—the mediating variables. In the present research, only the measure of social identity threat (i.e., mediating variable) was a measure of group-based threats that came before anti-immigrant attitudes (i.e., outcome variable). As such, we were not able to determine whether other group-based threats (e.g., symbolic threat) mediate the relationship between foreign language exposure and anti-immigrant sentiment. Future studies should measure symbolic threat, realistic threat, and epistemic threat prior to anti-immigrant attitudes to test them as mediators.
6. Conclusions
Our studies altogether suggest that exposure to a foreign language—regardless of whether they are consistent with Anglocentric constructions of American identity—lead White Americans to form less positive impressions of the targets and their conversation, sometimes experience an uptick in group-based threats, and display greater anti-immigrant attitudes. Moreover, their (in)ability to understand the conversation appears to dictate their perceptions of the immigrants and group-based threats. Considering how the U.S. is on its way to becoming a majority minority nation where foreign language speaking immigrants continue to grow and will comprise a significant portion of the U.S. population within the next several decades, it may be practical for White Americans to learn and become fluent in another language so as not to feel as threatened when they are exposed to a foreign language in a multicultural society. It is imperative to implement teachings and workshops focused on bringing awareness to the implications of being exposed to a foreign language, and inculcating individuals with a respect for multiculturalism and maintaining mother tongue. Bilingual and multilingual education should be widely implemented rather than discouraged, as this may help in shifting the narrative from “This is America, speak English!” to “This is America, feel free to speak any language you want!” Furthermore, future research should examine whether the present findings generalize to racial minorities—that is, do Black Americans, Asian Americans, and/or Latinx Americans experience group-based threats and display anti-immigrant attitudes to the same degree as White Americans when exposed to a foreign language?—and whether they are replicable in other Western countries also faced with impending demographic shifts. We hope our findings will encourage social psychologists to dig deeper into this issue and further examine how an exposure to a foreign language can influence one’s attitudes toward immigrants and group-based threats.
Author Contributions
Conceptualization, T.L. and L.E.M.; methodology, T.L. and L.E.M.; software, T.L.; validation, T.L. and L.E.M.; formal analysis, T.L.; investigation, T.L.; resources, T.L.; data curation, T.L.; writing—original draft preparation, T.L.; writing—review and editing, T.L. and L.E.M.; visualization, T.L.; supervision, L.E.M.; project administration, T.L. and L.E.M.; funding acquisition, T.L. All authors have read and agreed to the published version of the manuscript.
Funding
This research was funded by the KU Jack Brehm funds awarded to the first author.
Institutional Review Board Statement
The study was conducted according to the guidelines of the Declaration of Helsinki, and approved by the Institutional Review Board of the University of Kansas (STUDY00145106).
Informed Consent Statement
Informed consent was obtained from all subjects involved in the study.
Data Availability Statement
The data and materials used to conduct the research will be made available to any researcher for purposes of reproducing the results or replicating the procedure via email.
Conflicts of Interest
The authors declare no conflict of interest.
Footnotes
Note
1. The researchers note that the estimate for Asian immigrants is on the low end considering that the interviews were not available in an Asian language, and hence, the Asian respondents were necessarily fluent in English.
2. Although Newman et al. (2012) used the language subscale of the assimilationist threat scale (Paxton and Mughan 2006; e.g., “When in the company of Americans, immigrants need to speak to each other in English even if it is easier for them to use a common native language”) to tap into an epistemic component of cultural threat, the measure nonetheless fails to capture the processes that help explain why epistemic threat occurs—or specifically why immigrants should speak English when in the presence of other Americans—in the first place.
3. A total of 239 Whites who are not of Hispanic, Latino, or Spanish origin completed the study. However, 66 participants were removed for any of the following reasons: (a) they were not a U.S. citizen, (b) they failed the manipulation check, or (c) they spoke the language they were randomly assigned to—only for those who were assigned to one of the foreign language conditions, not the control condition.
4. Initially two separate three-items measures (i.e., positive perceptions of John and his parents and positive perceptions of the conversation) were used to measure positive perceptions of the family and their conversation, respectively. Results of the Pearson correlation indicated that positive perceptions of John and his parents was significantly highly correlated with positive perceptions of the conversation, r = 0.79, p < 0.001. For this reason, principal component analysis with direct Oblimin rotation was conducted, which yielded only one factor accounting for 80.83% of the total variance. Thus, the mean of all six items was computed to form a composite measure of participants’ positive perceptions of the family and their conversation.
5. We also ran a series of one-way ANCOVAs with participants’ household income, educational attainment, and political orientation as covariates. The findings were nearly identical in that the omnibus and Bonferroni post hoc tests were significant in a similar manner for each of the outcomes that were also significant in the analyses reported in Study 1.
6. A total of 462 Whites who are not of Hispanic, Latino, or Spanish origin completed the study. However, 138 participants were removed for any of the following reasons: (a) they were not a U.S. citizen, (b) they failed the one or both of the manipulation checks, or (c) they spoke the language they were randomly assigned to—only for those who were assigned to one of the foreign language conditions, not the control condition.
7. Initially, two separate three-items measures (i.e., positive perceptions of John and his parents and positive perceptions of the conversation) were used to measure positive perceptions of the family and their conversation, respectively. Results of the Pearson correlation indicated that positive perceptions of John and his parents was significantly highly correlated with positive perceptions of the conversation, r = 0.85, p < 0.001. For this reason, principal component analysis with direct Oblimin rotation was conducted, which yielded only one factor accounting for 85.68% of the total variance. Thus, the mean of all six items was computed to form a composite measure of participants’ positive perceptions of the family and their conversation.
8. We also ran series of two-way ANOVAs (language: Korean or German X subtitles: yes or no) and one-way ANCOVAs with participants’ household income, educational attainment, and political orientation as covariates. The findings were nearly identical in that the omnibus and Bonferroni post hoc tests were significant in a similar manner for each of the outcomes that were also significant in the analyses reported in Study 2.
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Figures and Tables
Enlarge this image.Figure 1. Mediation model for the relation between exposure to foreign language and anti-immigrant attitudes (Study 1). Note: Statistics for each path are unstandardized estimates. The unstandardized estimate in parentheses is the direct effect of foreign language on anti-immigrant attitudes, controlling for White Americans’ feelings of threat. * p < 0.05. *** p < 0.001.
Enlarge this image.Figure 2. Mediation model for the relation between exposure to foreign language with no subtitles and anti-immigrant attitudes (Study 2). Note: Statistics for each path are unstandardized estimates. The unstandardized estimate in parentheses is the direct effect of foreign language with no subtitles on anti-immigrant attitudes, controlling for White Americans’ feelings of threat. *** p < 0.001.
Mean scores and standard deviations for the main variables (Study 1).
| Variables | Korean |
Spanish |
German |
English |
One-Way ANOVA |
|---|---|---|---|---|---|
| Positive perceptions of the family and their conversation | 4.60 (1.06) a | 4.68 (1.25) a | 4.57 (1.27) a | 6.26 (0.73) b | F(3, 169) = 27.26 *** |
| White Americans’ feelings of threat | 3.60 (1.54) a | 3.87 (1.54) a | 3.93 (1.76) a | 1.69 (1.10) b | F(3, 169) = 23.75 *** |
| Anti-immigrant attitudes | 73.36 (26.25) a | 73.64 (28.82) a | 75.64 (21.53) a | 75.19 (23.90) a | F(3, 166) = 0.08 |
| Symbolic threat | 3.46 (1.55) a | 3.92 (1.72) a | 3.60 (1.49) a | 4.10 (1.64) a | F(3, 168) = 1.50 |
| Realistic threat | 3.25 (1.66) a | 3.81 (1.86) a | 3.48 (1.49) a | 3.72 (1.77) a | F(3, 169) = 0.95 |
| Epistemic threat | 3.45 (1.76) a | 3.50 (1.85) a | 3.43 (1.80) a | 3.52 (1.81) a | F(3, 169) = 0.02 |
Note: The F values represent the results of a one-way ANOVA to test for differences across conditions. Means within rows not having a common superscript differ at p < 0.05 using Tukey’s post hoc test. *** p < 0.001.
Table 2Mean scores and standard deviations for main variables (Study 2).
| Variables | Korean |
German |
Korean with Subtitles |
German with Subtitles |
English |
One-Way ANOVA |
|---|---|---|---|---|---|---|
| Positive perceptions of the family and their conversation | 4.38 (1.32) a | 4.23 (0.78) a | 6.01 (1.04) b | 5.76 (1.27) b | 5.93 (1.08) b | F(4, 318) = 38.85 *** |
| White Americans’ feelings of threat | 3.78 (1.52) a | 3.33 (1.36) ac | 2.55 (1.52) bd | 2.93 (1.55) bc | 1.90 (1.12) d | F(4, 318) = 17.88 *** |
| Anti-immigrant attitudes | 63.12 (27.75) a | 63.54 (23.65) a | 82.47 (17.74) b | 78.27 (20.75) b | 81.84 (19.79) b | F(4, 318) = 12.58 *** |
| Symbolic threat | 3.41 (1.56) a | 3.53 (1.28) a | 2.54 (1.28) b | 2.43 (1.23) b | 2.28 (1.08) b | F(4, 317) = 12.52 *** |
| Realistic threat | 2.94 (1.59) a | 2.77 (1.29) ab | 2.18 (1.25) b | 2.09 (1.31) b | 2.09 (1.17) b | F(4, 319) = 6.17 *** |
| Epistemic threat | 3.15 (1.77) a | 3.10 (1.46) a | 1.98 (1.12) b | 1.95 (1.21) b | 1.72 (1.10) b | F(4, 318) = 16.36 *** |
Note: The F values represent the results of a one-way ANOVA to test for differences across conditions. Means within rows not having a common superscript differ at p < 0.05 using Tukey’s post hoc test. *** p < 0.001.
© 2021 by the authors.