Abstract
Individuals are quickly categorized as members of a group (e.g., national, migrants) based on their standard or nonstandard accent. However, this has usually been studied in the context of the native language, and less is known about accent-based categorization when individuals speak a second language. In this research, we tested whether speaking in Portuguese or English would influence Portuguese participants’ evaluations and recall of Portuguese-accented and Brazilian-accented speakers. In Study 1, we investigated the perceived warmth and competence, the potentiality of forming relationships, and the assumed economic status of Portuguese-accented and Brazilian-accented men speaking in their native (Portuguese) or non-native language (English). Results indicated that Brazilians speaking English received the worst judgments compared to all other categories by Portuguese listeners. Study 2 used a Who Said What paradigm to test language and accent categorization in a different experimental setting, and predicted that Portuguese participants would be more likely to make errors when recalling Brazilian-accented men speaking in English. However, Portuguese-accented men speaking in English received the most errors (although, Brazilian-accented men speaking English were second). Overall results from this project provide insight into the nuanced effects of social categorization and evaluations of standard and nonstandard accented individuals speaking in their native and non-native languages in Portugal.
Everyone has accents. Accents give unique characteristics and personalities to individuals and groups of people that may be used to make inferences about others, such as their nationality, social class, sexual orientation, or age (see Fuertes et al., 2012 for a review). For instance, Portuguese individuals can quickly detect the accents of people from Lusophone (Portuguese-speaking postcolonial) countries with relative ease, leading to the rapid categorization of these individuals as minority members (i.e., immigrants) and facilitating voice-based discrimination (de Souza et al., 2016). Previous work mostly focused on categorization and evaluations of speakers when their native language was concerned, however, less work has focused on second languages. This research aimed to examine whether or not national accent-based categorizations, stereotyping, and interaction intentions may be influenced by speaking English in a Portuguese-speaking country.
Social Categorizations and Social Consequences
Categorization is defined as a process in which we may organize objects into meaningful categories with the underlying cognitive goal of simplifying the complexity of the world around us (Brewer, 1988; Hamilton & Trolier, 1986; Macrae & Bodenhausen, 2000). This process occurs rapidly and frequently in an automatic manner allowing us to attain membership in a number of social groups (Dovidio & Gaertner, 2010; Tajfel et al., 1971). The development of social categories, however, quickly leads to an “us versus them” mentality which may facilitate the preservation of harmful stereotypes leading to prejudice and discrimination against minority members (Dovidio & Gaertner, 2010; Fiske & Neuberg, 1990; Kang & Bodenhausen, 2015; Macrae et al., 1997; Tajfel et al., 1971; Turner, 1987).
According to the Stereotype Content Model, social stereotypes can be organized into two main dimensions: warmth and competence (e.g., Cuddy et al., 2011; Fiske et al., 2002, 2007). Warmth reflects whether a group is perceived as threatening or cooperative, whereas competence reflects whether the group is seen as capable or incompetent. These two dimensions create a quadrant, as warmth and competence are attributed independently, leading to different evaluations and emotions. Specifically, high warmth and high competence elicit admiration, high warmth and low competence trigger sympathy, low warmth and high competence elicit envy, and low warmth and low competence lead to contempt. Majority and minority groups (e.g., men and women, White and Black individuals; see Cheryan et al., 2015; Ellemers, 2018; Pedulla, 2014) are usually stereotyped along these dimensions, and their attribution of high/low competence and warmth is associated with different reactions.
Moreover, social categories and stereotypes are somewhat fluid in nature (Fiske, 2017; Stolier & Freeman, 2016), reacting to shifts across time and context (Garcia-Marques et al., 2006; Macrae & Bodenhausen, 2000), and therefore adding another layer of complexity to this process. Imagine a doctor who works in general practice. In the context of the workplace, the doctor might be perceived by their partner to be highly competent, friendly, or professional. If the doctor's partner saw them interacting with the other doctors at a bar for the first time, it is likely that the perceptions will shift in this specific context due to differences in perceived categorization (i.e., doctor versus bar-goer). Therefore, social categorization and perception are less related to the individual (or individual group), but rather, more related to the most prevalent characteristic of the individual (or individual group) in any given context (Fiske & Neuberg, 1990; Macrae & Bodenhausen, 2000; Moore-Berg & Karpinski, 2018). Moreover, individuals have the tendency to evaluate and/or react to the most prevalent social category as a method of attempting to efficiently categorize others (Fiske & Neuberg, 1990), but things quickly become complicated as individuals rarely belong to only one social category (Fiske, 2017; Moore-Berg & Karpinski, 2018).
Categorization of Multiple Minorities
Several theories have been proposed in an attempt to explain why and how we may perceive others with differing group memberships (i.e., cross-categorization), such as the additive model of social categorization (e.g., Crisp & Hewstone, 2000; Nicolas et al., 2017; Singh et al., 1997) and intersectionality models (e.g., Petsko et al., 2022). The additive model posits that ingroup or majority members evaluate others based on their “accumulation” of outgroup or minority memberships, with each additional category bringing more negative evaluations; in other words, each category acts as a sum of the whole. However, additive models typically do not account for nuanced or contextual findings and may sometimes, although at the basis of many concepts, be reductionary.
This additive approach is at the basis of the concept of double jeopardy (Beale, 1970; Blakemore & Boneham, 1994), stemming from feminist literature and stating that individuals belonging to two minorities may be subject to a “double hit” of prejudice. The original theory began by examining the experiences of Black women, who seemed to be perceived more negatively than both White women and Black men because of the intersection of racism and sexism (e.g., Beale, 1970; Blakemore & Boneham, 1994). However, since psychological research has advanced past the categorizations of solely race and gender, the notion of double jeopardy has been broadened to include any combination of multiple minority statuses (Browne & Misra, 2003; Hancock, 2007) to include age, disability, and sexual orientation in addition to gender and race (see Bowleg et al., 2003; Kang & Chasteen, 2009; Timmons et al., 2023). However, the effects of double or triple jeopardy are nuanced and have not received universal support, largely depending on the context or experimental design presented.
Altogether, although the additive models might be useful as a starting point for understanding categorizations of double minority categories, they can become an oversimplification of categorization and its effects because they are not applicable across contexts (Else-Quest & Hyde, 2016; Petsko & Bodenhausen, 2020, 2019). It is also unclear whether the final effects are due to the addition of each category or the interactions between each category in a given context, leading to different interpretations of the same results, and moreover, we cannot simply “add” the discrimination due to one minority category over that triggered by another minority group membership; for instance, we cannot simply add or multiply the effects of sexism and racism, and must instead opt for a more nuanced approach considering prototypicality and intersectionality (Goff et al., 2008; Hancock, 2007; Nicolas et al., 2017).
Categorization Through Intersectional Lenses
Intersectionality is a more consistent explanation of the assumptions that double jeopardy makes, stating that unique experiences and stereotypes may, or may not, be given to individuals with multiple minorities (e.g., Crenshaw, 1989; Ghavami & Peplau, 2012; Petsko et al., 2022). Similar to double jeopardy, it emerged from examining the experiences of Black women but takes context into account and acknowledges that, for double jeopardy of Black women to occur, both of the categories of race and gender must be made salient to the perceiver (Crenshaw, 1989). Intersectionality also acknowledges that all Black women do not experience the same types of discrimination due to unique experiences had by underlying categorical features of the individual such as social class (e.g., Crenshaw, 1989; Ghavami & Peplau, 2012; Wilkins, 2012). For instance, Wilkins (2012) found that individual economic status influenced the identities of Black women, whereas the wealthier individuals adopted more “racially-neutral” identities and were more befriended by White women than Black women from a low economic status. Intersectionality, therefore, must be assessed with all visible (or perceived) social categorizations through the lens of the perceiver in order to determine what is, or is not, considered to be prototypical (Billups et al., 2022; Sesko & Biernat, 2010).
The theory of intersectional invisibility assumes that sometimes, individuals do not receive discrimination based on the additions or interactions of their double minority status (i.e., double jeopardy or lens model), but instead, receive discrimination based on the unique qualities of their multiple subordinate identities which are neither positive nor negative in nature, particularly if they are perceived as nonprototypical of their constituent identities (Ghavami & Peplau, 2012; Purdie-Vaughns & Eibach, 2008; Sesko & Biernat, 2010). The authors built the theory based on the social concepts of androcentrism (e.g., male-oriented categorizations), ethnocentrism (e.g., ethnic majority-oriented categorizations which in Western contexts is typically White), and heterocentrism (e.g., straight-oriented categorizations), where the individual may be rendered invisible (i.e., “looked over”) on their absence of category membership, or in other words, made more nonprototypical. For instance, Hispanic men are not perceived as conceptually or prototypically similar to White men (e.g., not fitting ethnocentrism in White majority countries), and gay Hispanic men are not perceived as conceptually or prototypically similar to Hispanic and White men (e.g., not fitting ethnocentrism and heterocentrism) (Remedios & Snyder, 2018; Sykes, 2011). Experimentally, Sesko and Biernat (2010) found experimental evidence of this effect in which Black women were less recalled, and received more errors than Black men, White men, and White women in a memory recall paradigm in the United States.
Finally, the more recent lens model (see Petsko & Bodenhausen, 2020, 2019) proposes that, because of the nuances between individuals within a double minority category, we apply different lenses when perceiving outgroup targets based on what is most salient in any given context and adapt our perceptions accordingly. In other words, social categories may be made more, or less, salient based on situational aspects leading to perceptions of prototypicality or nonprototypicality (see also, stereotype activation/inhibition: Hogg, 2007; Macrae et al., 1995). Petsko and Bodenhausen (2019) found indeed that when gender, rather than race, was explicitly triggered, Black women became miscategorized with White women because of the larger cognitive emphasis participants ascribed to the targets’ gender; inversely, participants were more likely to associate men, and not women, with scientific attributes only when the lens of gender was applied. More closely aligned with the demographic groups of the current work, Reese and colleagues (2023) found that gay Brazilians did not receive hiring discrimination, but straight Brazilians did, only when competence was made salient in the experimental design; here, given the negative correlation between stereotypes of gay and Brazilians in the context of competence, the representation of one category (gay) overrides the stereotypical implications of the other category (Brazilians), and the stereotypes of being gay “canceled out” the stereotypes of being Brazilian (Carnaghi et al., 2022; Petsko & Bodenhausen, 2020, 2019).
The concept of intersectional prototypicality requires further investigation across contexts and paradigms, as this may produce nuanced results not easily predicated by simple additive models, namely depending on the (non)prototypicality effects of individuals with multiple memberships (Billups et al., 2022; Carnaghi et al., 2022; Ghavami & Peplau, 2012; Petsko & Bodenhausen, 2019; Purdie-Vaughns & Eibach, 2008) and on the fact that categories display a graded structure, with highly typical members tending to taken as the category default (Mervis & Rosch, 1981). Moreover, this model has not yet considered categorization based on vocal cues and languages.
Vocal Categorizations
One's voice and accent may give us more insight into where someone is (roughly) from (Dragojevic & Goatley-Soan, 2022; Rakić et al., 2011), their gender (Freeman & Ambady, 2011), or their race (Kurinec & Weaver, 2021) and sometimes, intangible attributes such as their sexual orientation (Fasoli et al., 2023) or social class (Campbell-Kibler, 2007). However, in a multilingual society, such perceptions could be changed by the language individuals speak as the language can activate different social categories or group representations. Imagine an individual might speak Portuguese at home with their partner and English at work with their colleagues; if their partner were to view them speaking English at work for the first time, it is possible that the perceptions surrounding the individual will shift in this specific context due to differences in perceived categorization (i.e., language spoken). One may argue this perception shift may occur due to a shift in the speaker's personality when speaking English compared to Portuguese (e.g., Chen et al., 2014; Dewaele, 2016), but importantly, the individual remains the same. This is particularly important as the second language is less associated with majority categories and therefore influences the categorization process (e.g., Paladino & Mazzurega, 2020; Rakić et al., 2011). As such, languages, and the accents we hold, are not solely vessels in which information may be passed, but rather provide a breadth of categorical information in their own right (Giles & Powesland, 1975; Pietraszewski & Shwartz, 2014).
These categorizations obtained through vocal information may, akin to the process of dissecting physical categories, result in the stereotyping and/or discrimination of others based on the stereotypical assumptions of their voice (Aronovitch, 1976; Fuertes et al., 2012; Kramer, 1964; Kurinec & Weaver, 2021; Stocker, 2017; Stoller et al., 2014), from as little as single-word utterances such as “Hello” (Baus et al., 2019). For instance, Kurinec and Weaver (2021) found that “sounding Black” (i.e., having an accent stereotypically prototypical of Black Americans) was associated with the attribution of negative characteristics (e.g., aggressive or uneducated) toward speakers, which mimics real-world effects of racially-coded accents that prompt some minority members to use more racially-neutral accents when speaking on the phone (and in other scenarios) to avoid potential discrimination (Gluszek & Dovidio, 2010). However, past research has shown that individuals are more prone to remember prototypically-congruent accents, which trigger stereotypical ideations of certain individuals (e.g., a German speaker narrating a sausage commercial), than prototypically-incongruent accents, which do not trigger these ideations (e.g., a German speaker narrating an olive oil commercial), perpetuating stereotypical categorizations and increasing the difficulty of individuals’ ability to forsake vocally-triggered stereotypes (e.g., Cartei et al., 2019; Fuertes et al., 2012; Hendriks et al., 2015; Rakić et al., 2011). This negativity toward nonprototypical individuals affects those who speak more than one language with an accent (and are unable to lose the accent of their native language). Paradoxically, learning a second language is both a sign of both higher- and lesser-educated individuals (Fuertes et al., 2012) where, for instance, native English speakers are praised by other English speakers for learning Spanish, but native Spanish speakers who speak English are generally perceived as uneducated, in comparison (e.g., Fuertes et al., 2012; Gluszek & Dovidio, 2010).
The Current Studies
Portugal and Brazil share a complex postcolonial history driven, in part, by overlapping aspects between cultures. In the Portuguese context, stereotypes and sentiments attributed to immigrants are generally aligned with the global [negative] perception of immigrants in which they may be more threatening, uneducated, or poorer than the nationals (Casquilo-Martins et al., 2022; Eaton, 1998; Guerra et al., 2015), and as Brazilians are the largest immigrant group in Portugal, these stereotypes are often attributed to them. Because of this, the Brazilian accent has not only become extremely salient to the Portuguese listener, but it has also been branded as a “less-educated” accent with numerous grammatical deviations from Portuguese from Portugal, which may facilitate discrimination against Brazilians living in Portugal (Casquilo-Martins et al., 2022). This implies that Portuguese listeners likely engage in stereotyping and prejudice when listening to Brazilian-accented speakers, causing Brazilians to sometimes speak in English so they are not perceived as being Brazilian although both Portuguese and Brazilians are native Portuguese-speakers (Casquilo-Martins et al., 2022; de Souza et al., 2016; Guerra et al., 2015). The language attitude literature clearly shows that nonstandard-accented speakers (e.g., foreign-sounding speakers) are perceived as less competent than standard-accented speakers (see Dragojevic et al., 2018).
Study 1 explored whether participants would hold negative evaluations of double minority individuals (and to a lesser degree, single-minority individuals) compared to other majority members (e.g., Beale, 1970; Blakemore & Boneham, 1994; Brewer, 1999; Crisp & Hewstone, 2000; Nicolas et al., 2017; Singh et al., 1997; Zhong et al., 2008) based on the concept of double jeopardy (e.g., Beale, 1970; Blakemore & Boneham, 1994). We tested this question when voice was the only cue providing information about nationality in a context-absent scenario, and additionally, assessed whether the language spoken (first vs second language) played a role in the impressions majority group listeners formed of speakers belonging to their group and to the minority group. Indeed, speaking a second language may activate the idea of a less prototypical group and emphasize the differences with the minority group. We considered four groups of speakers defined by their accent and language spoken: Portuguese men speaking Portuguese, Portuguese men speaking English, Brazilian men speaking Portuguese, and Brazilian men speaking English. In other words, Portuguese speaking Portuguese are the most prototypical group in Portuguese society (i.e., double majority members), followed by Portuguese speaking English (i.e., single majority members), Brazilians speaking Portuguese (i.e., single minority members), and Brazilians speaking English (i.e., double minority members). Because majority members (Portuguese individuals, in this case) may have more motivation to maintain positive majority identities than minority members (Leonardelli & Brewer, 2001; Moscatelli et al., 2017), and group differences may be made more salient than individual differences in the current paradigm (see Turner, 1987), we predicted Portuguese participants would be more likely to evaluate the [prototypically] farthest group (i.e., Brazilians speaking English) negatively based purely on their accents (see de Souza et al., 2016; Dragojevic et al., 2018; Rakić et al., 2011), supporting the theory of double jeopardy.
Study 2 sought to extend the findings of Study 1 by introducing a Who Said What paradigm (WSW) (Taylor et al., 1978) focusing on the recall of intersecting majority and minority group members. The WSW procedure examines a different psychological process (i.e., memory recall) than the one in Study 1 (i.e., attitudes), allowing us to test the assumptions of intersectional invisibility of the same members from Study 1. Specifically, the paradigm may influence different results as the paradigm itself is more sensitive to intergroup categorizations. That is, when introducing a memory recall paradigm, individuals may, or may not, perceive multiple minorities based on the combination of their social categories which may make them more, or less, prototypical to their main category and more, or less, susceptible to recall errors (e.g., Purdie-Vaughns & Eibach, 2008; Rosch, 1975). Being a less prototypical member of a majority group may lead to the individual being miscategorized (Rakić et al., 2011) or less likely to be recalled (Sesko & Biernat, 2010); in other words, rendered invisible (see Purdie-Vaughns & Eibach, 2008). Similar to prior research on double minority members (e.g., Brewer, 1999; Crisp & Hewstone, 2000; Nicolas et al., 2017; Rakić et al., 2011; Sesko & Biernat, 2010; Singh et al., 1997; Zhong et al., 2008) the base assumption is that individuals are more likely to correctly recall those from the same majority (Portuguese men speaking Portuguese, in this case), but at the same time, making more errors for those from double-minorities (i.e., Brazilian men speaking English). Embedded in this analysis is the assumption that double minority targets are poorly recognized by others and misattributed to others, relative to other target group members. Whereas this paradigm has traditionally been used to study the cross-sections of various physical attributes (see Bor, 2018; Klauer & Wegener, 1998), less work has been completed on intangible features, such as differences in vocal cues (e.g., Rakić et al., 2011).
Study 1
In addition to investigating potential effects of double jeopardy (e.g., Beale, 1970; Blakemore & Boneham, 1994), combined with the assumptions of (non)prototypicality (e.g., Billups et al., 2022; Sesko & Biernat, 2010), Study 1 also examined assumptions proposed by the stereotype content model (Cuddy et al., 2011; Fiske et al., 2002, 2007), which states that stereotypes of outgroup members generally converge onto the dimensions of perceived warmth and competence. We first predicted Brazilians speaking English (i.e., double minority members) would be evaluated as less warm (Hypothesis 1a [H1a]), less competent (Hypothesis 1b [H1b]), and less in potential closeness to participants (Hypothesis 1c [H1c]) by Portuguese individuals when compared to all other groups, positing that participants would accurately categorize group members and use this to discriminate against double minority individuals. Moreover, we predicted that the warmth and competence measures would be correlated with the potential closeness to participants across categories, given they investigate the same underlying principle (Hypothesis 2 [H2]). Finally, as Brazil is, and is socially perceived, as a poorer country than Portugal (Durante et al., 2013; Reese et al., 2024), we predicted that Portuguese participants would evaluate Brazilians, and specifically Brazilians speaking English, as having worse economic statuses than Portuguese speakers (Hypothesis 3 [H3]).
Method
Participants
Portuguese students were recruited through their university in exchange for course credit. We intended to recruit as many participants as possible and a G*Power sensitivity analysis (Faul et al., 2009) confirmed our sample size consisting of 51 individuals was sufficient to detect an effect (
Materials
Speakers
Speakers and phrases were pretested in a separate, working paper by the same authors. In the pretest (n = 72), 24 self-reported straight and gay men from either Portugal or Brazil were recorded in person in the lead author's city, and participants were tasked with evaluating how prototypical each speaker sounded of Portuguese, Brazilian, straight, and gay men. Importantly, to avoid sexual orientation as a confound, since it may introduce additional categorizations based on stereotypically-sounding straight or gay accents (Fasoli et al., 2023), we included only straight-sounding speakers in the current project (please note that the current study did not ask for categorization of sexual orientation). Specifically, we selected the two of the most prototypically-sounding straight Portuguese speakers and two of the most prototypically-sounding straight Brazilian speakers (in Portuguese). Each of the speakers uttered phrases in the Portuguese and English languages, resulting in eight total speakers for Studies 1 and 2. Importantly, all of the speakers were fluent in both Portuguese and English, and none of the participants realized that the speakers were used across both languages.
Sentences
All speakers recorded 48 neutral phrases adapted from previously pretested neutral items regarding masculinity from Cipriano et al. (2021), McDermott et al. (2019), and Plaza et al. (2017). Speakers recorded all of the 48 phrases in both Portuguese and English, of which 3 were shown to participants in the current project. The phrases were selected to keep the focus on the speaker, rather than the sentences’ meanings, by including common daily activities unrelated to stereotypes associated with any social group. Specifically, “I did some online shopping,” “I spent the afternoon at home,” and “I visited a new restaurant last weekend,” and were shown to participants in both Portuguese and English.
Procedure and Measures
Before the online experiment began, participants were urged to complete the study with headphones of good quality in a room without distractions. Following, participants were asked to form impressions about the personal characteristics of various speakers’ voices in different languages (i.e., Portuguese and English). Participants were told to give the first impression that came to mind, but, they were allowed to listen to the voice as many times as necessary. To avoid the possibility of guessing the intent of the study, either all English or all Portuguese speakers were randomly presented first to participants with a one-minute break separating each language. Within each language block, all speakers were randomized to account for any further issues.
Participants were given the option to report any technical difficulties in replaying the sound underneath each speaker's recording; most participants did not report any technical difficulties, and if there was a technical difficulty, that specific response was removed. Two separate Likert-type scales were presented to participants after listening to each speaker ranging from 1 (completely disagree) to 7 (completely agree) adapted from prior research including warmth and competence measures (i.e., Cuddy et al., 2011; Fiske et al., 2002; Pedulla, 2014). The first Likert-type scale included items related to warmth (i.e., “This person seems friendly,” “This person seems empathetic,” “This person doesn’t seem nice,” and “This person never helps their neighbors”), with the latter two being reverse-scored; the four items were combined to create the overall warmth measure and high reliability between items was found (Cronbach's α = .78). The next scale included items related to competence (i.e., “This person seems capable of learning new things,” “This person would be a good leader,” “This person does not seem intelligent,” and “This person seems lazy”) with the latter two being reverse-scored; the four items were combined to create the overall competence measure and high reliability between items was found (Cronbach's α = .82).
Next, participants were asked to estimate each speaker's monthly salary (by writing) and complete a Likert-type scale of the speakers’ economic status ranging from 1 (extremely low economic status) to 7 (extremely high economic status). Following, participants were asked to evaluate how probable it was they would engage in certain relationship-oriented activities with each speaker on a Likert-type scale ranging from 1 (not probable at all) to 5 (very probable). Instead of analyzing measures of closeness together, we argue that the dimension of competence may relate to the dimension of work closeness whereas the dimension of warmth may relate to the dimension of personal closeness. Specifically, work relationship formation was created by merging 3 items related to work competence (i.e., “How likely are you to … : work on an important project with this person?” “… ask this person for advice?” and “… trust this person to fulfill your tasks at work?”); high reliability between items was found (Cronbach's α = .86). Personal relationship formation was created by merging three additional items related to interpersonal warmth (i.e., “How likely are you to … : try to be friends with this person?” “… talk to this person about your life?” and “… invite this person to dinner at your house?”); high reliability between items was found (Cronbach's α = .83). Finally, participants were asked to report their demographics.
Results
Warmth and Competence
To evaluate measures of warmth and competence, a series of 2 (target nationality: Portuguese or Brazilian) × 2 (target language: Portuguese or English) repeated measures analysis of variance (ANOVA) tests were performed. An initial repeated measures ANOVA regarding perceptions of warmth revealed a significant main effect of target nationality, F(1, 50) = 8.61, p = .005,
Evaluations of warmth, competence, and closeness for target groups. Note. Speakers’ nationality is presented first with the spoken language in parentheses. Error bars represent 95% confidence intervals.
Evaluations of economic dimensions for target groups. Note. Economic status is situated on the primary (left) axis, whereas monthly salary is situated on the secondary (right) axis. Speakers’ nationality is presented first with the spoken language in parentheses. Error bars represent 95% confidence intervals.
A second 2 × 2 repeated measures ANOVA regarding perceptions of competence indicated a significant main effect of target nationality, F(1, 50) = 74.31, p < .001,
Closeness
The probability that participants would form either work or personal relationships with the speakers was evaluated in a 2 (target nationality) × 2 (target language) repeated measures ANOVA. An initial repeated measures ANOVA regarding the likelihood of forming work relationships with the speakers revealed a significant main effect of nationality, F(1, 50) = 50.75, p < .001,
Following, a 2 × 2 repeated measures ANOVA regarding the likelihood of forming personal relationships with the speakers indicated no significant main effect of nationality, F(1, 50) = 2.18, p = .146,
Correlations Between Warmth, Competence, and Closeness
We examined whether warmth would be correlated with personal closeness (probability of forming friendships) and competence would be correlated with work closeness (probability of forming work relationships). In fact, nearly all Warmth × Personal Closeness correlations were moderately strong for Portuguese speaking Portuguese (Pearson's r = .53, p < .001), Portuguese speaking English (r = .50, p < .001), Brazilians speaking Portuguese (r = .49, p < .001), and Brazilians speaking English (r = .39, p = .006). Alternatively, nearly all Competence × Work Closeness correlations were moderately strong for Portuguese speaking Portuguese (r = .37, p = .007), Portuguese speaking English (r = .46, p = .001), Brazilians speaking Portuguese (r = .49, p < .001), and Brazilians speaking English (r = .48, p < .001). Results supported H2.
Economic Measures
Again, a series of 2 (target nationality) × 2 (target language) repeated measures ANOVAs were conducted; an initial repeated measures ANOVA regarding perceived economic status found a significant main effect of nationality, F(1, 50) = 75.12, p < .001,
Lastly, a repeated measures ANOVA on estimated monthly salary found a significant main effect of nationality, F(1, 50) = 12.95, p < .001,
Discussion
The main hypotheses were supported by the findings. Brazilians speaking English received the most negative impressions and attitudes from participants, congruent with the additive model of cross-categorization and double jeopardy in which negative attributes are attributed to double minority members (e.g., Beale, 1970; Blakemore & Boneham, 1994; Crisp & Hewstone, 2000; Cuddy et al., 2011; Fiske et al., 2002; Purdie-Vaughns & Eibach, 2008; Singh et al., 1997), and the SCM which states that stereotypes converge onto the dimensions of warmth and competence (Cuddy et al., 2011; Fiske et al., 2002). Specifically, Brazilians speaking English were perceived, when compared to all other groups, the least warm, the least competent, and the least likely to benefit from work or personal relationships with participants. Importantly, speaking English did not act as a buffer against negative affects toward Brazilians, particularly toward measures regarding warmth. In this paradigm, where there is no context, and individual evaluation was present, simple effects of double minority status appeared to drive perceptions of Brazilians speaking English. This is, Brazilians speaking English were penalized in an additive manner, in which they did not receive benefits of positive stereotyping (i.e., being extroverted/high warmth (Sardinha, 2011)), but they did receive the downsides of negative stereotyping (i.e., being less educated/low competence, (Casquilo-Martins et al., 2022)). This effect was detected to a lesser degree regarding competence measures where Brazilians speaking in Portuguese were evaluated as less competent and less likely to benefit from work relationships than both Portuguese groups, although still more competent and more likely to benefit from work than Brazilians speaking in English. However, no interaction between Brazilian speakers was found when considering perceived economic status but it occurs when attributing salaries. While a Brazilian accent may have activated an overall impression of the minority's economic status (e.g., Duarte et al., 2013), salary was a more subtle measure allowing us to detect nuances of economic status and double jeopardy based on the language and accent of minority members, in which speaking English more negatively impacted ratings.
Study 2
Study 1 confirmed suspicions that Portuguese participants would evaluate double minority members (i.e., Brazilians speaking English) more negatively than other groups based on their voices. Study 2 attempted to expand upon these findings by considering intersectional categorization and prototypicality, making predictions based on prior Who Said What experiments (e.g., Bor, 2018; Klauer & Wegener, 1998; Rakić et al., 2011; Sesko & Biernat, 2010; Taylor et al., 1978), in which we predicted double minority status would drive memory recall errors. First, we predicted participants would have better accuracy toward targets speaking Portuguese than targets speaking English when guessing each target's nationality; as Portuguese is the first language of both target groups, we expected speaking English (i.e., increasing prototypical distance from the main groups) to worsen the overall hits, misses, and false alarms (H1). Next, we expected national and language categorizations to appear, resulting in more within-nation and within-language errors than between-nation and between-language errors, respectively (H2). We also predicted that Brazilians speaking English would receive the greatest amount of errors, being subject to the effects of multiple minority categorizations similar to the results obtained previously regarding multiple minority impression formation in Study 1 (H3).
Method
Participants
Portuguese students were recruited through their university in exchange for course credit. In total, there were 113 participants between 18 and 50 years old (M = 21.46, SD = 6.68), and a G*Power sensitivity analysis (Faul et al., 2009) confirmed our sample size consisting of 113 individuals was sufficient to detect an effect (
Materials
The speakers in Study 2 were the same as those included in Study 1. However, each speaker spoke four neutral phrases related to a specific theme (e.g., food or daily activities) to implement the Who Said What procedure and fit the cover story. In total, 24 neutral phrases were used in Study 2. In addition, 28 faces were selected from the Chicago Face Database (CFD; Ma et al., 2015) based on how neutral they appeared to be of both Portuguese and Brazilian men and were examined in a pretest including 12 Portuguese and Brazilian natives (M = 32.17, SD = 11.11, six men and six women), with the eight most neutral faces used in the main experiment. Each face was then paired with one of the speakers for the entirety of the experiment.
Procedure
Before the experiment began, participants were urged to complete the study with headphones of good quality in a room without distractions. Following, participants were told they would hear parts of a conversation between eight individuals and a researcher (whose aim in their study was to understand more about locals’ daily activities); the individuals in the conversation would respond in the language they use most often at work (i.e., Portuguese or English), producing the four target groups. Participants were asked to form impressions about the group of individuals interviewed, although the real main task was memory recall. Before continuing on to the main experiment, participants were given an opportunity to test their headphones (with a recording of piano music).
Once prepared, participants were asked to listen to the edited conversation with the interviewer in which the interviewer's questions were always written, instead of spoken (i.e., “What did you do yesterday afternoon?”); to avoid contamination, the interviewer also did not have a photo. After each section's question was presented, the paired voices and faces of each speaker were randomly shown to participants to avoid pattern memorization and/or recall. After listening to all eight speakers respond to all four questions), participants were then asked to complete a distractor task to avoid recency effects; specifically, participants were asked to list as many animals as they could within a 90-s time frame. The next section did not appear until the time was finished.
Afterwards, participants were shown either a new or old phrase and were asked to decide if the statement was new or old. Importantly, all of the new phrases were created from the old phrases to increase the overall difficulty of recall; for instance, “I spent the afternoon in the garden” was changed to, “I spent the afternoon at a museum.” Additionally, as both Portuguese and English languages were used in this paradigm, new and old sentences were shown in the language it was spoken, or created from, although the instructions were always in Portuguese. If the participant decided the statement was new, the next statement would immediately appear. If the participant decided the statement was old, even if the statement was actually new, the participant would be asked to select which speaker said the phrase, adding multiple levels of cognitive processes.
Results
Hits, Misses, and False Alarms
To frame subsequent results and provide a better picture of the processes at hand, the amount of hits (i.e., correct [speaker] guesses), misses (i.e., selecting an old phrase as new), and false alarms (i.e., selecting a new phrase as old, then selecting a speaker) were used as separate dependent variables in a series of 2 (target nationality) × 2 (target language) repeated measures analysis of variance (ANOVA) tests. An initial 2 × 2 repeated measures ANOVA regarding the total number of hits (i.e., selecting the correct speaker for each statement) revealed a significant main effect of nationality, F(1, 112) = 5.77, p = .018,
A second 2 × 2 repeated measures ANOVA regarding misses revealed a significant main effect of nationality, F(1, 112) = 27.07, p < .001,
A final 2 × 2 repeated measures ANOVA regarding the false alarms found no significant main effect of nationality, F(1, 112) = 1.72, p = .193,
Within- Versus Between-Errors
Categorization based on nationality and language was first examined by evaluating all four error types (i.e., within-nation/within-language, within-nation/between-language, between-nation/within-language, and between-nation/between-language) in a 2 (Nationality error: within or between) × 2 (Language error: within or between) repeated measures ANOVA. Importantly, the responses for within-nation/within-language errors were multiplied by two for all targets as there was only one incorrect target for this type of error; in contrast, all other errors had two incorrect targets. The repeated measures ANOVA found a significant main effect of nationality error, F(1, 112) = 4.11, p = .045,
Errors Involving Brazilians Speaking English
As Brazilians who spoke in English consistently received the worse evaluations in Study 1, similar findings were predicted to unfold when comparing Brazilians speaking English versus all other target groups; specifically, all four error types were compared with each target group in a 2 (nationality error: within or between) × 2 (language error: within or between) × 4 (target group) repeated measures ANOVA, although the focus of analysis was the interaction between all three variables. As predicted, there was a significant interaction between all three variables, F(3, 336) = 5.64, p < .001,
Types of errors within each target group. Note. Speakers’ nationality is presented first with the spoken language in parentheses. Types of errors are represented by the following abbreviations: within-nation/within-language (W-N/W-L), within-nation/between-language (W-N/B-L), between-nation/within-language (B-N/W-L), and between-nation/between-language (B-N/B-L). Error bars represent 95% confidence intervals.
Within-Nation/Within-Language Errors
A planned contrast found participants were more likely to confuse Portuguese speakers with one another (M = 1.33, SD = 1.80) than Brazilian speakers with one another (M = 1.03, SD = 1.41), t(1330.53) = −3.04, p = .002, d = .17. Against predictions, Brazilians speaking English (M = 1.08, SD = 1.54) did not receive the most within-nation/within-language errors compared to all other groups in a three versus one comparison, t(1330.53) = −1.14, p = .256, d = .03. In fact, Portuguese speaking English (M = 1.36, SD = 1.78) received the most within-nation/within-language errors when compared to all other groups in a three versus one comparison, t(1330.53) = 2.17, p = .030, d = .06.
Within-Nation/Between-Language Errors
An additional planned contrast supported the main hypothesis, indicating participants were more likely to confuse Brazilians speaking Portuguese with Brazilians speaking English (M = .63, SD = .82) than Portuguese speaking Portuguese with Portuguese speaking English (M = .30, SD = .57), t(1330.53) = 3.40, p < .001, d = .19. Moreover, a three versus one planned contrast found Brazilians speaking English (M = .74, SD = .87) had significantly more within-nation/between-language errors than all other groups, combined, t(1330.53) = 3.31, p < .001, d = .18.
Between-Nation/Within-Language Errors
Participants were more likely to confuse Portuguese and Brazilian targets speaking English (M = 1.26, SD = 1.02) than Portuguese and Brazilian targets speaking Portuguese (M = .79, SD = .86), t(1330.53) = 4.74, p < .001, d = .25. Unexpectedly however, Brazilians speaking English (M = .95, SD = .84) did not receive the most between-nation/within-language errors compared to all other groups in a three versus one comparison, t(1330.53) = −.88, p = .380, d = .02; instead, Portuguese speaking English (M = 1.57, SD = 1.19) received statistically more errors than all other groups in a three versus on comparison, t(1330.53) = 6.35, p < .001, d = .17.
Between-Nation/Between-Language Errors
A final planned contrast found participants were less likely to confuse Portuguese speaking Portuguese with Brazilians speaking English (M = .32, SD = .68) than confusing Portuguese speaking English with Brazilians speaking Portuguese (M = .80, SD = .90), t(1330.53) = −2.60, p = .010, d = .14. In fact, Brazilians speaking English (M = .17, SD = .48) actually received the least errors compared to all other groups in a three versus one comparison, t(1330.53) = −4.55, p < .001, d = .13, whereas Portuguese speaking English (M = .91, SD = .99) again received the most errors compared to all other groups in a three versus one comparison, t(1330.53) = 4.13, p < .001, d = .23. Results from all error types largely did not support H3.
Discussion
Study 2 provided information on memory of majority and minority members depending on their accent and language spoken. First, Portuguese speaking Portuguese (i.e., the majority group) received the most accurate responses when considering hits, misses, and false alarms, meaning participants made more errors for English speakers which was predicted by past findings of the WSW paradigm. Importantly, more within-nation errors were made (compared to between-nation errors), signaling that the categorization of speakers’ nationalities was used even when nationality was not explicitly triggered (as in Study 1). However, this effect appeared to be stronger when targets spoke in Portuguese (i.e., participants’ native language) considering participants were better at correctly guessing Portuguese-speaking targets than English-speaking targets, and as a result, targets speaking in a non-native language (i.e., English) were typically subject to higher error rates, supporting the hypothesis that being prototypically farther from a group influences greater recall errors. Brazilians speaking English received more within-nation/between-language errors than all other groups, meaning participants were likely to confuse Brazilians speaking English with Brazilians speaking Portuguese. Portuguese speaking English received the highest between-nation/within-language and between-nation/between-language errors, meaning participants were more likely to confuse them with Brazilians speaking English and Brazilians speaking Portuguese, respectively. At the same time, Portuguese speakers speaking English were less likely to be correctly guessed than Brazilians speaking English. Therefore, whereas Brazilians did not fully lose their categorization as Brazilian when speaking English, Portuguese did appear to lose their categorization as Portuguese when speaking English, becoming also vulnerable to poorer recognition and miscategorization (see Rakić et al., 2011). In other words, when Portuguese speak English (in a group context), they may be perceived as being a foreigner by other Portuguese speakers. These results do go against the initial hypotheses made by prior WSW results (e.g., Rakić et al., 2011; Sesko and Biernat, 2010), and highlight the need for additional research using different contexts and prototypes, especially in non-English languages and cultures when language is primed.
General Discussion
Social and vocal categorizations are generally quick in everyday life and drive the attitudes we may have toward individuals from certain social groups. Due in part to the complex nature of cross-categorization, there have been several models and theories proposed aimed at better understanding this process, some of which are less-tested in experimental manners. In this project, we explored the potential effects of double jeopardy on double minority members (e.g., Beale, 1970; Blakemore & Boneham, 1994), predicting that members with two minority statuses would receive the most negative evaluations (e.g., Purdie-Vaughns & Eibach, 2008; Rosch, 1975). Specifically, we examined how Portuguese individuals evaluated and recalled Portuguese or Brazilian men speaking in either native Portuguese or non-native English. The following discussion should be viewed through the lens of this specific context of majority-minority interactions and perceptions.
Study 1 first looked at the stereotypical assumptions a majority group might hold toward other majority group members (i.e., Portuguese speaking Portuguese), single majority group members (i.e., Portuguese speaking English), single minority members (i.e., Brazilians speaking Portuguese), and double minority members (i.e., Brazilians speaking English) in accordance with the stereotype content model (e.g., Cuddy et al., 2011; Fiske et al., 2002, 2007), past research on double minority members (e.g., Beale, 1970; Blakemore & Boneham, 1994), and predictions of the additive model of cross-categorization (Crisp & Hewstone, 2000; Singh et al., 1997). Importantly, Study 1 was performed in a context-absent scenario in which participants were simply told to evaluate each speaker based on their voice where no contextual information was provided. We found that Brazilians speaking English were evaluated more negatively than all other target groups (apart from the general economic measure), supporting predictions made by the additive model of cross-categorization and the effects of double jeopardy. More specifically, Portuguese participants viewed Brazilians speaking in English as the least warm, least competent, and least likely to form work or personal relationships (with themselves) compared to all other groups, including Brazilians who spoke Portuguese. In line with the additive model and double jeopardy, it appears that speaking English did not buffer negative stereotypical attitudes participants held toward Brazilians and instead acted as the opposite, driving participants’ negative perceptions. Participants seemed to be aware of the targets’ intersected majority/minority statuses, resulting in the negative evaluation of the prototypically farthest group from the majority group (i.e., Portuguese individuals).
This effect did not, however, predict participants’ evaluations of the general economic statuses of each group; Portuguese speakers were perceived as having higher economic statuses than Brazilian speakers, regardless of language spoken. As argued by Duarte and colleagues (2013), individuals may internalize the objective economic statuses of nations and apply these perceptions to the citizens of those nations (although, these perceptions will never be true as each nationality has wealthy and nonwealthy citizens). For instance, although Germans may be perceived as wealthier than Indians (Reese et al., 2024), one may of course encounter wealthier Indians and poorer Germans. This psychological bias was evident in Study 1 suggesting that language does not affect majority members’ ideations of minority members when considering the general economic statuses of individuals (although, language does affect other outlets). Still, when asking about salaries, the double jeopardy effect was observed suggesting that salaries may capture stereotyping and biases that take into consideration the nationality and language intersection. This may indeed relate to assumptions on types of jobs people do, for instance, that Brazilians speaking English may work in low-paying jobs where speaking English is required (e.g., tourism industry or domestic work) and plays on the idea that learning English as a foreigner paradoxically equates to less education (e.g., Fuertes et al., 2012; Gluszek & Dovidio, 2010).
Study 2 implemented a Who Said What? paradigm (e.g., Bor, 2018; Klauer & Wegener, 1998; Rakić et al., 2011; Sesko & Biernat, 2010; Taylor et al., 1978) thus adding contextual triggers to the experimental design and further exploring the possibility of double minority discrimination in a memory-focused experiment with group recall instead of individual evaluation. Specifically, participants were told each speaker was asked to respond in the language they most often use at work, both signaling and increasing the saliency of the language spoken (i.e., Portuguese or English). We posited that participants would make more recall errors for Brazilians speaking English (i.e., double minority group) based on assumptions made from double jeopardy and intersectional invisibility, but findings were not as straightforward. Portuguese participants were more likely to correctly guess other Portuguese speakers (resulting in less misses and false alarms) than all other groups only when Portuguese speakers spoke in Portuguese; this was expected, as majority group members should be more likely to detect and recall other majority group members according to past WSW and prototypicality research on categorization effects (Rosch, 1975). At the same time, more within-nation errors than between-nation errors were made, suggesting participants accurately categorized not only the majority groups, but also the minority groups, as well (although this effect was lessened when targets spoke in English).
However, against initial predictions, Portuguese speaking English received more errors and less correct guesses than Brazilians speaking English, suggesting that whereas Brazilians somewhat lost their national categorization when speaking English, Portuguese individuals mostly lost their categorization when speaking English. At face value, this outcome diverges from the results of Study 1, predictions of double jeopardy (e.g., Beale, 1970; Blakemore & Boneham, 1994), and predictions made by the additive model of categorization which requires that perceivers are able to first be able to correctly perceive both present categories (Nicolas et al., 2017; Petsko & Bodenhausen, 2019). Cross-categorization is highly variable and difficult to interpret across contexts so findings should not be accepted at face value. Instead, the findings appeared to be guided more by assumptions of intersectionality and nonprototypicality, in which different, and intersecting identities, may produce unexpected outcomes for single and double outgroup members (e.g., Ghavami & Peplau, 2012; Goff et al., 2008; Purdie-Vaughns & Eibach, 2008; Sesko & Biernat, 2010).
Interpreting results through this more nuanced view of categorization allows us to examine the context in which the experiment was conducted and the actors involved. First and foremost, speaking English may influence paradoxical perceptions among perceivers and speakers where it can be considered a positive or negative trait (Fuertes et al., 2012; Gluszek & Dovidio, 2010). Because of this, Brazilians in Portugal may sometimes speak English in an attempt to avoid the negative stereotypes of speaking Brazilian Portuguese (e.g., less educated) and not be perceived as being Brazilian, largely in contexts where accent-based discrimination and xenophobia may exist, such as social services or work (Casquilo-Martins et al., 2022; de Souza et al., 2016; Guerra et al., 2015). Simply put, Brazilians sometimes hide their Brazilian identity by speaking English in Portugal. Portuguese individuals, however, do not need to hide their identity in Portugal by speaking English. Importantly, the results of Study 2 highlight that Portuguese speakers speaking English were not categorized as Portuguese and were perhaps miscategorized as Brazilian (i.e., the most between-nation/within-language errors) because of the paradigm disrupt that is being a native-speaking Portuguese speaking English to other Portuguese individuals (e.g., Fuertes et al., 2012; Gluszek & Dovidio, 2010). To the Portuguese participant, speaking English seemed to equate with being foreign. If we then consider Portuguese speaking English as miscategorized and conceptually the same as Brazilians speaking English (i.e., an “honorary” double minority in this context), the results from Study 2 then converge with the results of Study 1 (and prior WSW experiments involving double minorities and intersectional invisibility (e.g., Bor, 2018; Klauer & Wegener, 1998; Rakić et al., 2011; Sesko & Biernat, 2010; Taylor et al., 1978), and fits with predictions made by double jeopardy (e.g., Beale, 1970; Blakemore & Boneham, 1994). Another important aspect to consider is the lens of the experiment.
Since language was explicitly triggered, and national categorization was implicitly triggered by accent, we argue that Brazilians were perhaps made more similar (to one another), where Portuguese were made less similar (to one another) (see Petsko & Bodenhausen, 2019); that is, participants might have conceptualized both Brazilian groups as Brazilian independently of the language spoken, resulting in less overall saliency for these speakers when considering the group as a whole. The opposite might have occurred for Portuguese individuals who, speaking English, were viewed as prototypically further than Portuguese individuals speaking Portuguese, resulting in greater saliency for the single majority group and the increase of errors made by participants. In any case, speaking English as a Portuguese or Brazilian individual in a group setting may cause negative impressions toward those members to arise and influence overall memory or attribution errors, suggesting that although English is a “global language,” speaking Portuguese may yet be better for individuals from Portuguese-speaking countries. More specifically, when the focus is on negative biases (Study 1), people may scrutinize the double minority group which represents the least prototypical group (i.e., the group deviating the most from the norm). However, when the focus is on group memory/categorization (Study 2), it may be that people look more at the more prototypical group, rather than the nonprototypical group, which may further explain why Brazilians speaking English did not have as strong as an effect as Portuguese speaking English.
Despite the theoretical advancements this project provides, there are a few limitations worth noting. First and foremost, each speakers’ language fluency was not explicitly tested, although this may undoubtedly lead to confusing categorizations; instead, each speaker was evaluated on how prototypical their accent appeared to be of either Portuguese or Brazilian speakers. Hypothetically, the results from Study 1 might be justified by the possibility that all of the Brazilians speaking in English had a lower level of English (compared to the Portuguese individuals), rather than the effects of stereotyping related to cross-categorization. Similarly, regional Portuguese and Brazilian accents were also not considered, widening the potentiality of missed categorizations based on specific accents found within each nation (although, in fact, all speakers were pretested to have a prototypical accent respective of their nationality). Finally, specific linguistic nuances may have been weakened or eliminated entirely in the WSW paradigm used as speakers uttered phrases independently of other actors, which may have resulted in specific target groups being perceived as more, or less, warm or competent than others. Future research could better adjust for these discrepancies, recruiting participants from the same level of language fluency or a specific regional upbringing, and at the same time, opting for a more natural replication of conversation. Moreover, future conclusions may benefit greatly from a minority versus majority point of view (i.e., Brazilian participants’ evaluations of Portuguese and Brazilian speakers).
Majority groups have sociopolitical power over minority groups, generally, so interdisciplinary work comparing not only majority-minority relations, but also minority-majority relations are imperative due to globalization. It is increasingly important to understand the effects of one's accent on how they may be perceived by others, especially when speaking in another language. Although plausible that speaking English may benefit certain individuals in an international setting, results suggest the opposite may also be true. Speaking English not only influenced negative stereotypical perceptions of Brazilians, but also led to greater categorical errors for Portuguese individuals (and to a lesser degree, for Brazilians), highlighting the nuanced effects of cross-categorization triggered by vocal clues. Importantly, the results from this work do not propose that all Brazilians or Portuguese individuals who speak English will receive negative perceptions or recalls, but instead, the results suggest that, from a Portuguese point of view, using Portuguese participants in Portugal, these effects may occur in some contexts based on theories of categorization and intersectionality. If we conducted the same study from a Brazilian point of view, using Brazilian participants in Brazil, we might receive the results that Portuguese speakers speaking English were worst-rated in Study 1, for instance, which only means these nuanced effects must be continued to be researched for theoretical and practical applications across contexts. Finally, the results are contingent on whether or not any of the aforementioned categories are correctly categorized in the first place—if a category is not correctly categorized or perceived, due to a number of factors, models quickly fall apart and require more nuanced interpretations. These results may be regarded as disheartening by some, but in fact, they provide valuable information into how majority groups conceptualize individuals from one or more minorities.
Footnotes
Acknowledgments
Author’s Note
Jonathan Reese is also affiliated at HEI-Lab, Universidade Lusófona, Lisbon, Portugal.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The principal author of this project received a research grant from the Fundação para a Ciência e a Tecnologia (FCT) (Reference: PD/BD/150702).
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Use of AI
AI was not used in any part of this manuscript, data collection, or analysis.