Toward a Century of Language Attitudes Research: Looking Back and Moving Forward

Documentation

The predominant and longest-standing line of research has focused on documenting people’s attitudes toward different language varieties and the speakers who use them; this research is primarily descriptive. Past studies have documented people’s attitudes toward a wide range of linguistic variation, including different languages (e.g., Lambert et al., 1960); regional (e.g., Kinzler & DeJesus, 2013), ethnic (e.g., Rodriguez et al., 2004), and social-class accents and dialects (e.g., Giles, 1970); foreign accents (e.g., Lindemann, 2003); code-switching (e.g., Genesee & Bourhis, 1982); gay and lesbian speech (e.g., Fasoli & Hegarty, 2020); powerful/powerless language (e.g., Gibbons et al., 1991); lexical formality (e.g., Levin et al., 1994) and diversity (e.g., Bradac & Wisegarver, 1984); as well as individual linguistic variables (e.g., Labov, 1966).

Collectively, this research shows that language varieties within a given society can be ordered on a hierarchy of prestige, typically corresponding to the socioeconomic status of the social groups they are associated with. Varieties associated with socioeconomically dominant groups tend to carry high prestige; these typically include majority group languages, standard varieties—namely those that have been codified (Milroy & Milroy, 1999)—and other forms associated with powerful groups (e.g., straight speech)—hereafter collectively referred to as high prestige varieties. In contrast, varieties associated with socioeconomically subordinate groups tend to carry low prestige; these typically include minority group languages, nonstandard varieties—namely those that diverge from codified norms, including most regional and ethnic dialects and foreign accents—and other forms linked to stigmatized groups (e.g., gay/lesbian speech)—hereafter collectively referred to as low prestige varieties.

Cross-culturally and worldwide, speakers of low prestige varieties are typically rated less favorably on status traits than speakers of high prestige varieties (Fuertes et al., 2012; Preston, 1989). Some low prestige varieties are denigrated more than others (Coupland & Bishop, 2007; Lindemann, 2005), and the more a person’s speech diverges from high prestige forms, the less status they tend to be attributed (Dragojevic et al., 2017). This evaluative downgrading can occur in both formal (e.g., school) and informal (e.g., home) contexts, but is typically accentuated in the former (Creber & Giles, 1983). It also tends to emerge regardless of listeners’ own group affiliations; when differences do emerge, they tend to be differences in degree of preference rather than direction of preference (Woolard & Gahng, 1990). Indeed, speakers of low prestige varieties themselves often ascribe low status to members of their own linguistic communities (Kinzler & DeJesus, 2013) and, sometimes, may even exaggerate it, reflecting what has been termed the “minority group reaction” (Lambert et al., 1960) and “linguistic insecurity” (Labov, 1966).

Despite this, speakers of low prestige varieties are sometimes rated more favorably on solidarity traits than speakers of high prestige varieties by members of their own linguistic community (Luhman, 1990) or more widely (Kinzler & DeJesus, 2013). This covert prestige (Labov, 1966), however, does not extend to all low prestige varieties (Coupland & Bishop, 2007) nor does it extend necessarily to all speakers of a given variety (Gallois et al., 1984; Luhman, 1990). Indeed, low prestige varieties—especially foreign accents—are often downgraded on both dimensions (Cramer, 2016).

Explanation

The documentation literature has provided clear evidence that language attitudes occur in nonrandom ways. In an attempt to explain these empirical regularities, another line of research has focused on identifying the causal mechanisms that underlie language attitudes. In this way, the documentation literature—although itself primarily descriptive—has served as an important springboard for theoretical development (see Giles & Marlow, 2011).

In their seminal work, Lambert et al. (1960) reasoned that language attitudes reflect social group stereotypes. Consistent with this proposition, extant research shows that evaluative reactions to language are often based on learned stereotypic associations rather than any inherent properties of language itself (Giles & Niedzielski, 1998). Members of different social groups speak differently (Labov, 1966). People become aware of these associations at a young age and use language to infer speakers’ social group memberships (e.g., ethnicity, social class; Kinzler et al., 2010). Although social categorization can be based on any socially diagnostic cue, spoken language often assumes primacy over other cues. For instance, both adults and children are more likely to categorize others based on accent than race-related visual cues (Kinzler et al., 2009; Rakić et al., 2011a).

Once a social category becomes salient, associated stereotypes are activated and influence person perception (Dragojevic, 2018). Stereotype content typically reflects the socio-structural relations between groups in society (Fiske et al., 2002). Status stereotypes are based on perceived socioeconomic status (Ryan, 1983). Consequently, speakers categorized as belonging to socioeconomically subordinate groups (e.g., speakers of low prestige varieties) are typically attributed less status than speakers categorized as belonging to socioeconomically dominant groups (e.g., speakers of high prestige varieties). Solidarity stereotypes are based on ingroup loyalty (Ryan, 1983) and perceived competition (both realistic and symbolic: Kervyn et al., 2015). Consequently, speakers categorized as belonging to one’s linguistic ingroup and noncompetitive outgroups (e.g., some regional accented-speakers) are typically attributed more solidarity than speakers categorized as belonging to competitive outgroups (e.g., most foreign-accented speakers). The tendency to ascribe high solidarity to members of one’s own linguistic community is especially pronounced among people who identify strongly with that community (Luhman, 1990) and perceive it to have increasing vitality (i.e., status, demographics, institution support; Ryan et al., 1984).

Language-based social categorization and stereotyping are prone to considerable variation, for several reasons. First, most linguistic cues mark multiple identities, at various levels of specificity (see also notion of indexical fields: Eckert, 2008). For instance, a New York accent marks a local identity (New York), a regional identity (East Coast), and a national identity (American), among others. Which identity becomes salient depends on listeners’ familiarity with the associations in question (Ryan, 1983) as well as the social comparative context in which social categorization occurs (Abrams & Hogg, 1987). Second, at all levels of specificity, listeners can make errors. Errors are more likely to occur for more specific social categories (e.g., Korean vs. Chinese) than abstract categories (e.g., native vs. nonnative; Lindemann, 2003). This variation in categorization can be consequential. To the extent that different social categories are associated with different stereotypes, then the same speaker can be evaluated differently depending on how they are categorized (Dragojevic et al., 2018; Yook & Lindemann, 2013), with consequences for how listeners subsequently perceive and process their speech (McGowan, 2015; Rubin, 1992). Third, in addition to inferring which groups speakers belong to, listeners also use language to gauge how prototypical of those groups speakers are (Dragojevic et al., 2017) which, in turn, can influence the extent to which speakers are stereotyped (Blair et al., 2002). Fourth, listeners’ perceptions of the socio-structural relations between different groups in society—and their stereotypes of those groups—can vary as a function of listeners’ own group affiliations. Consequently, even though different groups of listeners may categorize a speaker identically, they may evaluate the speaker differently (Coupland & Bishop, 2007; Gallois et al., 1984).

Neither the activation of a specific social category nor knowledge of specific linguistic stereotypes, however, is necessary to evoke language attitudes. For instance, infants express a clear social preference for native- over foreign-language speakers, without any knowledge of specific linguistic stereotypes (Kinzler et al., 2007). This preference is likely based on familiarity. Related, people may form value judgments about specific linguistic features (e.g., double negative are illogical), without necessarily having any knowledge of how those features are distributed socially. In this way, language may elicit evaluative reactions in a more direct fashion (see Roessel et al., 2018).

Language attitudes can also be influenced by listeners’ processing fluency, defined as the subjective ease or difficulty listeners experience processing a person’s speech (Dovidio & Gluszek, 2012). Listeners’ processing fluency can be influenced by various factors, including speakers’ language (Cristia et al., 2012); environmental factors, such as background noise (Munro, 1998); and listeners’ own characteristics, such as their familiarity with the variety in question (Gass & Varonis, 1984) and their social dominance orientation (Hansen & Dovidio, 2016). Disruptions in fluency—regardless of source—can negatively bias listeners’ evaluations of speakers (Dragojevic & Giles, 2016; Roessel et al., 2019), and these effects persist even after controlling for categorization and stereotyping (see fluency principle: Dragojevic, 2020; Dragojevic et al., 2017).

Listeners’ expectations can also play a role. Prior to hearing a person speak, listeners may form expectations about the person’s likely language behavior, typically based on contextual norms and/or social group stereotypes (Burgoon, 1993). Compared to speakers who confirm listeners’ expectations, speakers who negatively violate expectations (i.e., use a less prestigious variety than expected) engender more negative evaluations, whereas speakers who positively violate expectations (i.e., use a more prestigious variety than expected) engender more positive evaluations (for norm-based violations, see Creber & Giles, 1983; for stereotype-based violations, see Hansen et al., 2018).

Development

Given overwhelming empirical evidence showing that language attitudes are socially mediated—and therefore potentially prone to variation across the lifespan—researchers have also focused their attention on the developmental trajectory of language attitudes, examining when, how, and from whom language attitudes are acquired. This research shows that perceptual sensitivity to linguistic variation begins at birth. Newborns can distinguish between their maternal language and foreign languages (Mehler et al., 1988). By the time they are 5 months old, infants look longer at linguistic ingroup than linguistic outgroup members and can distinguish between native and foreign accents (Kinzler et al., 2007), as well as some regional varieties of their maternal language (Butler et al., 2011). The first clear social preferences based on language emerge as early as 10 to 12 months after birth. For instance, infants at this age are more likely to accept toys from native over foreign language speakers (Kinzler et al., 2007).

This preference for linguistic ingroup members continues into early childhood and extends to other domains. Preschoolers preferentially select native-language and native-accented speakers as friends and trust them more over foreign-language and foreign-accented speakers, respectively (Kinzler et al., 2007, 2011). Preschoolers also begin to coordinate linguistic and social knowledge. For instance, they assume that people who speak an unfamiliar foreign language are more likely to be of a different race, dress differently, and live in a different style of housing than people who speak their native language (Hirschfeld & Gelman, 1997). They also begin to make language-based judgments about others’ personality traits (e.g., how smart and nice they are; Kinzler & DeJesus, 2013). Initially, these judgments continue to favor linguistic ingroup members, even if they are associated with negative stereotypes (Day, 1982). However, by the first few years of elementary school, most children endorse the same linguistic stereotypes as adults: They evaluatively upgrade on relevant dimensions speech they associate with positive stereotypes, regardless of whether it is their own (Kinzler & DeJesus, 2013).

Children can acquire specific linguistic stereotypes through exposure to both overt and covert messages about language. Overt messages explicitly assign value to language (e.g., double-negatives are illogical) and its users (e.g., people who use double-negatives are uneducated). Such messages are prevalent in the educational system, where high prestige varieties typically function as the code of instruction and are explicitly prescribed and promoted as the “best” and most “correct” ways to speak (Milroy & Milroy, 1999). Equally overt messages can also come from parents, peers, and other community members across various contexts (Marlow & Giles, 2010).

Covert messages implicitly link value to language. For instance, the visibility and salience of different varieties in the public sphere—or the linguistic landscape (Landry & Bourhis, 1997)—can lead people to make inferences about the value and power of those varieties and the groups they are associated with (Dailey et al., 2005). Media portrayals contribute to these perceptions. For instance, standard speakers are overrepresented in the media, whereas nonstandard speakers are underrepresented, which can undermine the perceived status of the latter (Lippi-Green, 2012). Moreover, media portrayals of different linguistic groups are stereotypical: Standard speakers tend to be portrayed in positive roles, whereas nonstandard speakers—particularly foreign-accented speakers—in negative roles. Such stereotypical portrayals have been documented in Disney movies (Lippi-Green, 2012), children’s cartoons on cable and network television (Dobrow & Gidney, 1998), and American primetime television (Dragojevic et al., 2016). Repeated exposure to these stereotypical portrayals can contribute both to the formation and maintenance of language-based stereotypes (Mastro, 2009).

Consequences

Complementing the documentation literature described earlier, researchers have also afforded attention to the social and behavioral consequences of language attitudes. Whereas documentation studies treat language attitudes as the key output of interest, studies focused on consequences treat language attitudes as socially meaningful input that influences subsequent responses. Stated differently, language attitudes are assumed to mediate the effects of linguistic variation on these more distal outcomes, though admittedly this assumption is only rarely tested explicitly (e.g., Fasoli & Hegarty, 2020; Roessel et al., 2019).

Language attitudes have been shown to influence language learning and maintenance (Gardner, 1982), as well as language choice during interpersonal and intergroup encounters (Bourhis, 1984). They have also been shown to influence nonlanguage behaviors and other social judgments across a range of applied contexts. Collectively, this research shows that speakers of low prestige varieties frequently face prejudice and discrimination (see Gluszek & Dovidio, 2010b; Roessel et al., 2020). For instance, nonstandard speakers tend to be judged as less suitable than standard speakers for high status employment (e.g., manager; Giles et al., 1981; Rakić et al., 2011b). This effect tends to be more pronounced the stronger a speaker’s nonstandard accent is (Carlson & McHenry, 2006) and can emerge regardless of speakers’ actual qualifications (Roessel et al., 2019). Similarly, gay- and lesbian-sounding speakers are judged as less suitable for leadership positions than straight-sounding speakers because they are attributed less status (Fasoli & Hegarty, 2020). In educational settings, students often rate nonnative-accented instructors less favorably than native-accented instructors on teaching quality and effectiveness (Gill, 1994; Subtirelu, 2015)—even when they themselves are nonnative speakers (Hendriks et al., 2018)—and may actively avoid courses taught by nonnative-accented instructors (Rubin & Smith, 1990). Educators may also display prejudice and make more negative inferences about the academic performance and future prospects (e.g., happiness) of nonstandard-speaking students (Choy & Dodd, 1976).

In simulated criminal proceedings, nonstandard speakers are often judged as more guilty than standard speakers of both blue-collar (e.g., assault) and white-collar crimes (e.g., fraud; Dixon et al., 1994, 2002), and as more likely to be re-accused of a crime, regardless of the quality of evidence presented against them (Dixon & Mahoney, 2004). They also tend to be judged as less credible, truthful, and accurate eyewitnesses (Frumkin, 2007). Moreover, nonstandard speakers are typically perceived as less persuasive (Dragojevic et al., 2020; Tsalikis et al., 1991) and are less likely to secure compliance than standard speakers (Bourhis & Giles, 1976; Giles et al., 1975). The former effect, however, may be attenuated or even reversed when nonstandard speakers advocate positions incongruent with listeners’ expectations (Giles et al., 1995).

Change

In light of the wide range of adverse consequences negative language attitudes can have for those involved, researchers have also devoted some effort to developing and assessing the effectiveness of various interventions designed to counteract those negative effects. Although research in this area is relatively scarce, both macro-institutional and individual-based interventions appear to have promise. Some studies have examined how government language policies influence language attitudes over time. This research shows that such policies can have significant effects on the vitality of different linguistic groups in society (Bourhis, 2019) which, over time, can enable language attitude change, with status stereotypes typically more difficult to change than solidarity stereotypes (Genesee & Holobow, 1989; Woolard & Gahng, 1990).

Other research in this area consists of one-off studies testing the effectiveness of specific (typically idiosyncratic) interventions at improving language attitudes. For example, educating people about their biases via dialect awareness training (Bozoglan & Gok, 2016; Wolfram & Schilling, 2015) and/or explicitly instructing them to not let their biases influence their evaluations (Roessel et al., 2019) both appear to attenuate the evaluative downgrading of foreign-accented speakers. Inducing people to engage in perspective-taking with respect to foreign-accented speakers—either through imagined (Weyant, 2007) or direct experience speaking a foreign language (Hansen et al., 2014)—also appears to be effective. Likewise, structured intergroup contact with foreign-accented speakers (Kang et al., 2014) and interventions which increase the ease with which listeners process foreign-accented speech (Dragojevic, 2020) have also had success. Other interventions have focused on promoting an inclusive group identity. For instance, inducing participants to recategorize outgroup regional-accented speakers in terms of a common ingroup identity increases listeners’ sense of connection with those speakers and promotes more favorable evaluations (Abrams & Hogg, 1987; Dragojevic & Giles, 2014). Interestingly, imitating an unfamiliar regional accent also appears to reduce bias, perhaps by facilitating perspective-taking and/or making a common ingroup identity salient (Adank et al., 2013).

Summary

As the foregoing attests, people make a wide range of judgments about others simply based on how they speak (documentation). Those judgments emerge early in life (development) and are the product of various causal mechanisms (explanation). They can have a wide range of effects, including prejudice and discrimination (consequences), some of which can be attenuated through various interventions (change).

Approach

Different disciplines have studied language attitudes using different perspectives and methodologies. On the one hand, this reflects the multidisciplinary nature of this topic and has given rise to a diverse and robust literature. On the other hand, it has resulted in the creation of disciplinary and methodological silos, which rarely communicate with one another. Moving forward, an interdisciplinary approach could help break down some of these barriers and further propel theoretical development, as well as promote social change. For instance, adopting a historical perspective may allow us to better understand how and why attitudes toward some varieties have changed over time (e.g., due to policies, norms; see Genesee & Holobow, 1989; Woolard & Gahng, 1990) and, hence, provide additional insight about why other varieties continue to be negatively evaluated today (see Hegarty, 2020). Similarly, better integration of the five lines of research described above would be fruitful. For instance, increased attention to the causal mechanisms underlying language attitudes (explanation) can facilitate the design and implementation of more effective interventions (change), which, in turn, can help reduce language-based prejudice and discrimination (consequences).

Methodology

Integrating new methodological approaches with theories would also help disentangle the language attitudes processes. For instance, a growing body of research is now using the voice averaging approach—which involves voice morphing—to test the impact of stereotypical speech on person perception (see Kachel et al., 2018). This approach may be useful to further test the role of social categorization and stereotyping in the language attitudes process (see Blair et al., 2002). Related, the use of big data could help us document how language attitudes are expressed in a wider range of contexts (e.g., social media; Durham, 2016), as well as provide us with a better understanding of how language attitudes change over time. More longitudinal research would also be helpful in this respect. While we know that specific interventions are effective at promoting language attitude change, it remains unclear to what extent those changes persist over time.

More consistency in the way studies approach language attitudes could also be beneficial. For instance, studies relying on vocal stimuli often do not measure listeners’ perceptions of speakers’ social group memberships, making it difficult to interpret findings, compare results across studies, and fully understand underlying processes (e.g., categorization, stereotyping). In light of this, future studies utilizing vocal stimuli should measure and report listeners’ perceptions of speakers’ social group memberships whenever possible (see Preston, 1989). This—coupled with the development of additional ‘best practice guidelines’—would not only provide researchers with a standard procedure to follow but would also facilitate study comparison and replicability.

Intersectionality

More attention should also be devoted to intersectionality (see Levon, 2015). People belong to many different social categories, each of which may be conveyed by one or more vocal and nonvocal cues; how the intersectionality of these different identities influences social categorization and speaker perceptions remains understudied in the language attitudes literature (but see Rakić et al., 2020). For example, work by Campbell-Kibler (2011) has shown that, when vocal cues associated with different identities intersect, listeners form complex impression of speakers and may privilege some vocal cues over others when judging certain characteristics. It could be interesting to examine whether speakers whose voice simultaneously indexes multiple stigmatized identities (e.g., gay, ethnic minority) are subject to stronger discrimination than speakers whose voice indexes only a single stigmatized identity. Considering all these aspects in tandem is particularly important for the design and implementation of successful, theoretically-grounded interventions.

Target Perspective

Past research has primarily focused on the listener’s perspective, documenting how people react to others’ speech. Gluszek and Dovidio (2010a) have advocated for more focus on the speaker’s perspective, examining how speakers—particularly those who use low prestige varieties—perceive and evaluate their own speech. Indeed, nonstandard accented speakers expect to be stereotyped and discriminated because of their accent (Derwing, 2003). Similarly, gay and lesbian speakers who believe they sound gender-atypical expect to be categorized as gay by others (Fasoli et al., 2018). This not only has detrimental effects on speakers’ psychological well-being and social functioning (Birney et al., 2020; Gluszek & Dovidio, 2010b; Lippi-Green, 2012), but also implies that speakers need to navigate stigma, or at least stigma expectations. One way speakers try to cope with stigma is by modulating their speech to either communicate or conceal their social identity (Giles, 2016). This has been observed among both nonstandard-accented (Moyer, 2007) and gay speakers (Daniele et al., 2020). However, such behavior—whether conscious or unconscious—is potentially stressful. Hence, interventions should focus not only on reducing discrimination toward speakers of low prestige varieties, but also on helping to improve the well-being of those speakers who have to cope with and navigate prejudice and discrimination.