Abstract
Mindreading is integral to navigating social interactions, especially when they involve people from diverse cultural backgrounds who may have different ways of thinking, knowledge, and beliefs. The present study aimed to explore how accent as a social group membership cue impacts our perspective-taking abilities. To do this, 566 participants took part in the Strange Stories task wherein they listened to stories told by either native- or foreign-accented protagonists. Half of the stories required participants to take the protagonists’ perspectives (“mindreading” stories). The other half of the stories were void of mental states to capture participants’ baseline capacities for speech and narrative comprehension (“general comprehension” stories). On both types of stories, participants who listened to native-accented protagonists performed better than those who heard foreign-accented protagonists. Through examining the predictors of mindreading, we found that participants’ perspective-taking performance when listening to native-accented protagonists was simply explained by their general comprehension ability. Better mindreading of foreign-accented protagonists, however, was linked not only to comprehension but also with an intergroup factor: namely, having fewer friendships with people of different races and accents. We discuss the unexpected directionality of this association in light of how accents may impact the ways in which we engage with people who speak differently to ourselves via the social meaning attached to them.
Keywords
Introduction
In today’s globalized societies, we often come across people of diverse cultures. Whether it be the way we look (e.g., race), act (e.g., customs), or speak (e.g., accent), various social group membership cues of cultural background impact intergroup interactions (for reviews, see Kinzler, 2021; Moore-Berg & Karpinski, 2019). One socio-cognitive factor theorized to be crucial to the way we interact is mindreading (Astington, 2003). Mindreading refers to our capacity to deduce mental states in others, such as their thoughts, beliefs, or emotions and, in turn, explain and predict their behavior (Wellman, 2001). While mindreading has traditionally been examined in interpersonal contexts (for reviews, see Hughes & Leekam, 2004; Weimer et al., 2021), a handful of studies have applied an intergroup lens. These studies suggest that our effectiveness in taking others’ perspectives may depend on the cultural group membership of those we are interacting with, at least when we are explicitly made aware of cultural similarities and differences (Ekerim et al., 2020; Gönültaş et al., 2019; Perez-Zapata et al., 2016; Sudo & Farrar, 2020; Todd et al., 2011). Most often in our day-to-day lives, however, we spontaneously infer people’s backgrounds without explicit exchange of culturally specific information (Weatherhead et al., 2018). The present study, therefore, investigated mindreading when cultural group membership is simply gleaned from the way one speaks—namely, via one’s accent.
Cultural Group Membership and Mindreading
In one of the earliest studies on intercultural mindreading, Todd et al. (2011) proposed that we tend to impose our own perspectives on people from our own culture, whereas differences between ourselves versus people from other cultural groups encourage us to look beyond our own experiences and infer their mental states. To test this theory, Todd et al. (2011) presented German undergraduate students with a false-belief mindreading task wherein they manipulated the protagonist’s culture: half of the participants read about a German protagonist (identified with a German name and depicted with light skin and hair), and the other half read about a Turkish protagonist (identified with a Turkish name and dark skin and hair). Participants watched the protagonist hide an object in one container and leave the scene. While the protagonist was away, the second character moved the object to another container. At this point, half of the participants were given explicit information on where the second character moved the object to (“the red container”), whereas this information was left ambiguous for the other half (“another container”). When the protagonist returned to the scene, participants were asked where the protagonist will look for the object.
Compared to participants in the German condition who read about a protagonist from their own cultural group, those in the Turkish condition were less likely to respond erroneously based on their own knowledge of the object’s actual location. Instead, they were more likely to take the protagonist’s perspective to accurately predict that the protagonist will look in the original location. This was only the case, however, for participants who were explicitly told where the object was moved to. That is, the effect of cultural group membership on mindreading was found specifically when the participants had to put aside their own belief and respond based on the protagonist’s mental state. Based on these findings, Todd et al. (2011) concluded that differences in cultural group membership prime effective perspective-taking by suppressing egocentric responding (see also Sudo & Farrar, 2020, for similar findings with children as young as four years of age).
In cases where one’s perspective is not in conflict with that of the interlocutor, however, similarities in cultural group membership may instead facilitate mindreading by giving a common understanding of the situation. Researchers have explored this possibility using the Strange Stories task (White et al., 2009), which does not require participants to reconcile their own perspective conflicting with that of the protagonist—instead, success in the task is achieved via integrating the perspectives of the characters within each vignette. Furthermore, in addition to the “mindreading” stories that require perspective taking, the Strange Stories task presents participants with vignettes that are void of mental states. These “general comprehension” stories serve as a measure of participants’ baseline ability to follow narratives and integrate multiple information to reason accordingly.
In one study that used the Strange Stories task, Perez-Zapata et al. (2016) presented Australian undergraduate students with vignettes that featured protagonists of two different cultures: half of the participants read stories about protagonists from Australia and the other half read about protagonists from Chile—a country culturally dissimilar to Australia. Consistent with Todd et al. (2011), cultural group membership was manipulated by the name and race of the protagonist, but further emphasized by referring to the protagonists’ birthplace, as well as cultural activities and surrounding objects. Participants who read about protagonists from their own culture scored higher on the mindreading stories than those presented with protagonists from another culture. The effect of cultural group membership was not found in stories that assessed participants’ general comprehension ability. Perez-Zapata et al. (2016) speculated that cultural similarity may have promoted shared understanding of social situations that facilitated perspective taking (see also Ekerim et al., 2020; Gönültaş et al., 2019).
Predictors of Intercultural Mindreading
To empirically test the role of perceived similarity on intercultural mindreading, Ekerim et al. (2020) presented Turkish adults with the Strange Stories task that featured protagonists from their own culture (Turkey) or one of two other cultures (Syria or Norway), along with a measure of perceived similarity. Additionally, Ekerim et al. (2020) administered measures of prejudicial attitudes to investigate the possibility that these attitudes may preclude perspective taking. Due to its geographic proximity, Syrian people are often perceived as culturally similar to Turkish people; however, many Turkish people hold prejudicial attitudes against Syrians as a result of the influx of Syrian refugees (Yitmen & Verkuyten, 2018). In contrast, Norway is largely culturally dissimilar to Turkey, but Turkish people typically are not prejudiced against Norwegians given the lack of political conflict between these countries.
Consistent with Perez-Zapata et al.’s (2016) findings, Ekerim et al. (2020) found that participants who took the perspective of protagonists from their own culture (Turkey) scored higher on the Strange Stories task compared to those who read about protagonists from the two other cultures (Syria and Norway). Although participants were equally poor at mindreading people from the two cultures that are not their own, differences were found in measures of perceived similarity and prejudicial attitudes. As expected, Turkish participants rated Syrians to be more similar to themselves than the Norwegians, but they expressed more prejudice towards Syrians compared to Norwegians. Moreover, these intergroup attitudes were associated with mindreading scores. When perceived similarity was controlled for, prejudice was found to predict poor performance on reading the minds of protagonists from other cultures.
Taken together, there is growing evidence that cultural group membership impacts mindreading. On one hand, cultural differences may encourage efforts to put aside one’s own perspectives and understand others’ mental states (Sudo & Farrar, 2020; Todd et al., 2011). On the other hand, cultural differences that elicit prejudicial attitudes may dissolve efforts required for perspective taking, resulting in poorer mindreading of people from other cultural groups (Ekerim et al., 2020; Gönültaş et al., 2019; Perez-Zapata et al., 2006). The methodological manipulation of cultural group membership in these studies, however, calls into question the possibility that the effect of culture on mindreading was found because participants were primed to perceive the protagonists from other cultures as “different”. Namely, cultural group membership was explicitly highlighted through culturally specific names, racial features, unique cultural practices, geographical origin, and in some cases with statements that emphasized cultural dissimilarities (e.g., “wears clothes that’s very different from yours”; Sudo & Farrar, 2020). When we come across people from other cultures in the real world, cultural differences are typically not overtly asserted to this extent; instead, they are more often inferred.
Accent-Based Cultural Group Membership
One implicit cue to cultural group membership is the accent we speak with. From as early as the preschool years, children infer that people who share their native accent are from their own country (Weatherhead et al., 2018) and expect them to have similar cultural practices (e.g., food, clothing) to people of their own community (Wagner et al., 2014). Accent-based inferences become increasingly nuanced and entangled with intergroup attitudes into adulthood, wherein accents come to represent their respective cultural groups and, in turn, elicit stereotypes and prejudicial attitudes associated with those groups (for reviews, see Fuertes et al., 2012; Giles & Rakić, 2014; Imuta & Spence, 2020). Although the indirect nature of accent as a cultural group membership cue often permits accent-based prejudice to go unnoticed (Gluszek & Dovidio, 2010), the profound implications of discrimination against foreign-accented speakers are evident in various real-world contexts (e.g., employment; for a review, see Spence et al., 2022).
To the best of our knowledge, the effect of accent-based cultural group membership on mindreading has yet to be examined in adults. Nevertheless, two studies have been conducted with children. In Witt et al.’s (2022) study, 4-year-old German children took part in a false-belief task wherein the protagonist spoke German with either a native, German accent or a foreign, Polish accent. Accent was the only cue to cultural group membership in this study, and Witt et al. (2022) found that children performed similarly on the mindreading task regardless of the protagonist’s accent. In a more recent study by Anderson et al. (2023), 4- and 6-year-old children from New Zealand were presented with a visual perspective-taking task by two researchers: one who spoke English with a native, New Zealand accent and another with a foreign, American accent. In Anderson et al.’s (2023) study, accent was coupled with additional information about the speakers’ geographical origin and cultural knowledge. Consistent with Witt et al.’s (2022) study, however, accent-based cultural group membership did not impact children’s performance on the mindreading task.
On one hand, these findings may suggest that accent is not a meaningful enough cue for cultural group membership to influence mindreading. On the other hand, there may be methodological and developmental explanations for the null results. In both Witt et al. (2022) and Anderson et al. (2023), the participants were preschoolers, but the interlocutors were adults—that is, they were all “outgroup” members in terms of age. The effect of accent-based group membership may have been overridden by preschoolers’ sensitivity to age-based group membership (Seehagen et al., 2018). Additionally, accent-based group membership may not have been made salient enough. In Witt et al. (2022), children were only exposed to the protagonist’s accent for 15 seconds prior to the false-belief task. In Anderson et al.’s (2023) study, the “foreign” accent (American) was likely familiar to children given the popularity of American children’s entertainment in New Zealand (Wright, 2025). Preschoolers categorize familiar and own accent variants together (Jones et al., 2017; McCullough et al., 2017), thus, children in Anderson et al.’s (2023) study may not have perceived New Zealand- versus American-accented speakers as belonging to meaningfully distinct cultural groups. Finally, accent may not impact preschool-aged children’s mindreading because they have yet to acquire relevant intergroup attitudes that modulate perspective taking (Ekerim et al., 2020; Gönültaş et al., 2019). Accent-based group membership may only influence mindreading from late childhood onwards when accents first begin to elicit prejudicial sentiments attached to their respective cultural groups (for a review, see Imuta & Spence, 2020).
The Present Study
The present study aimed to investigate the effect of accent-based cultural group membership on mindreading in adults and how this is influenced by intergroup factors. By adulthood, beliefs regarding the meaningfulness of accents as a social evaluation cue are well-established (Hansen, 2020). Adults readily use accents to identify the speaker’s cultural group membership and, in turn, draw inferences about the speaker’s traits based on intergroup attitudes associated with their cultural group (for reviews, see Fuertes et al., 2012; Gluszek & Dovidio, 2010). For these reasons, we hypothesized that adults would be better at taking the perspectives of native-accented speakers from their own cultural group compared to foreign-accented speakers from other cultural groups.
Specifically, we tested Australian undergraduate students—a sample wherein the effect of cultural group membership on mindreading was previously observed, at least when culture was manipulated via explicit identification of culturally specific names, race, birthplace, and cultural preferences (Perez-Zapata et al., 2016). To assess mindreading, we used the Strange Stories task and manipulated the protagonists’ accent-based cultural group membership. Participants listened to vignettes narrated by either protagonists who spoke English with a native, Australian accent or with foreign, Malaysian and Singaporean accents. We chose these foreign varieties because their linguistic properties are distinctive of Asian accents (Cavallaro et al., 2020; Hashim, 2020), but English is an official language in both countries. Thus, many Malaysians and Singaporeans acquire English side-by-side with other languages (Thirusanku & Yunus, 2012) and are fluent in English; this is especially the case for those living in Australia (Australian Bureau of Statistics, 2021). These foreign accent varieties, therefore, allowed us to isolate accent from common linguistic confounds, such as fluency of speech. Protagonists who spoke with Malaysian- and Singaporean-accented English jointly represented the “foreign” accent group, given that “Singaporean Malaysian English” is traditionally classified as a single entity due to their shared linguistic origin and overlap in linguistic features (Ling, 2020).
In line with previous studies that have examined the predictors of intercultural mindreading (Ekerim et al., 2020; Gönültaş et al., 2019), we investigated how perceived similarity is linked with perspective taking. Since the introduction of multicultural policies, Australia has welcomed a large number of immigrants from Asia, many of whom are from Malaysia and Singapore (Australian Bureau of Statistics, 2021). Despite this, the Australian identity remains “grounded in the idea of separateness from Asia” (p. 37) and “whiteness” together with English competence continue to be key characteristics of this identity; thus, Asian immigrants are generally regarded as “different others” (p. 23) (Martin et al., 2023). In the context of the present study, therefore, our Australian participants were expected to perceive the foreign, Malaysian- and Singaporean-accented protagonists as less similar to themselves compared to the native, Australian-accented protagonists. We predicted that lower perceived similarity ratings, in turn, would be linked to poorer mindreading.
In addition to perceived similarity, we examined prejudicial attitudes as predictors of intercultural mindreading. Due to the Australian government’s commitment to multicultural policy, overt racial-ethnic discrimination is frowned upon by many in society (for a review, see Ben et al., 2024), yet subtle sentiments of prejudice persist (Forrest et al., 2021). For example, Asian immigrants in Australia often experience social exclusion, bullying, and discrimination based on their foreign accents (Dovchin, 2020; Dryden & Dovchin, 2024). To capture implicit biases against immigrants, we used a measure of “modern” prejudice (Akrami et al., 2000) and assessed perceptions on how far immigrants should assimilate and integrate into Australian culture (Nshom & Kahlimzoda, 2020). We also asked participants how easily they thought accents can be changed (accent stability) and to what extent accents are predictive of the speaker’s traits (accent diagnosticity), as these beliefs are linked to prejudicial attitudes toward foreign-accented individuals (Hansen, 2020). In line with previous findings (Ekerim et al., 2020; Gönültaş et al., 2019), we expected participants who harbor negative attitudes toward immigrants and foreign-accented speakers to be worse at taking the perspectives of these people.
Lastly, we examined another intergroup factor that has yet to be explored in relation to intercultural mindreading. In particular, we asked participants to report the amount of exposure and friendships they have with those who do not share their race or accent. Intergroup contact has been found to powerfully shape intergroup attitudes (for a recent meta-analysis, see Van Assche et al., 2023), especially when it takes the form of friendships that involve high-quality interactions (e.g., Davies et al., 2011). With over 30% of the current population born outside of Australia and nearly a quarter speaking a language other than English at home (Australian Bureau of Statistics, 2021), Australia affords intergroup contact. Nevertheless, the spread of immigrant populations is disproportionate across cities and regional areas; thus, the level of exposure to cultural and linguistic diversity that Australians experience differs largely depending on one’s region of residence. Those who live in areas where there are greater opportunities for contact with immigrant populations, in turn, have been found to display less anti-immigrant sentiments (Guan & Pietsch, 2022). Extending on the idea that prejudicial attitudes hinder mindreading (Ekerim et al., 2020), for the first time, we tested the possibility that those who are less frequently exposed to and have fewer friends who do not share their race or accent will be worse at understanding the mental states of “foreign” others.
Method
Participants
A total of 832 first-year undergraduate psychology students at The University of Queensland, Australia, took part in this study for course credit (see supplementary material for details on sample size decision). Of these participants, 246 were excluded due to taking part in the study twice (n = 16); having a clinical diagnosis which could impact performance on the key variables of interest in this study (autism spectrum disorder, dyslexia, hearing impairment; n = 17); failing the attention checks (n = 109); and failing the accent manipulation check (n = 104). Given that the key manipulation in this study pertained to accent-based group membership, we did not specify accent in the eligibility criteria for our study to avoid bringing undue attention to this variable in the recruitment process. To ensure that the foreign accents featured in this study (Malaysian and Singaporean) were not ingroup accents for any of our participants, however, we excluded data from those who reported to speak with a Malaysian or Singaporean accent (n = 20). Of the remaining 566 participants who contributed data to the analyses (Mage = 20.55 years, SD = 3.97; 373 women, 193 men), 450 reported to speak English with a native, Australian accent and 116 identified with a non-Australian accent variety. 1 Approximately 60% identified their ethnicity as Australian, and the majority of other participants reported various Asian ethnicities (see Tables S1 and S2 in the supplementary material for a full breakdown of accents and ethnicities). The study was approved by The University of Queensland Human Research Ethics Committee.
Measures
Strange Stories task
An adapted version of the Strange Stories task (White et al., 2009) was used to test participants’ ability to take the perspective of others who speak with a native or foreign accent. In the standard Strange Stories task, participants are presented with written vignettes about protagonists conveyed from a third-person perspective. In our study, to manipulate the protagonists’ accents, we modified the story scripts to be told from the protagonists’ first-person perspective and had native, Australian-accented speakers and foreign, Malaysian- and Singaporean-accented speakers narrate the stories as the protagonists. Thus, participants in the Native Accent condition heard the protagonists tell their stories in a native, Australian accent, and those in the Foreign Accent condition heard the protagonist tell their stories in foreign, Malaysian and Singaporean accents. Furthermore, to ensure that the only cue to the protagonists’ cultural group membership was their accent, we removed any culturally specific names, objects, and activities from the story scripts (see supplementary material for scripts).
We presented the vignettes as short video clips featuring the auditory accent narrations, accompanied by subtitles and relevant illustrations. We created two sets of video clips for each accent condition: one featuring four women, and the other with four men (see supplementary material for details on the speaker selection process). Participants listened to speakers of their own gender to avoid gender being a confounding social group membership cue. Each participant listened to sixteen vignettes: eight mindreading stories and eight general comprehension stories. The sixteen vignettes were presented in a randomized order. At the end of each vignette, participants were asked to provide a written response to a question that assessed their understanding of the mental states (mindreading) or information unrelated to mental states (general comprehension) represented in the story. The general comprehension stories in the original Strange Stories task were designed to capture participants’ baseline ability to make sense of narratives and reason accordingly (White et al., 2009). Given the auditory nature of our task, in our study, these stories also served to tap into how well participants understood the protagonists’ speech at the linguistic level.
Participants’ responses on each question were scored on a scale of 0 to 2, with higher scores representing greater mindreading or general comprehension. To attain the maximum score of 2 on a given question, participants had to provide a coherent explanation for the outcome of the story via integrating two key pieces of information. For mindreading stories, participants needed to display their understanding of both: (i) the protagonist’s mental state (e.g., belief that they would be arrested) and how this was motivated by (ii) the mental state of the second character (e.g., knowledge that the protagonist stole an item). For general comprehension stories, participants needed to demonstrate their causal reasoning of how (i) one situation (e.g., an umbrella falling in a spiky bush) was linked to (ii) a second situation (e.g., holes in the umbrella). Responses that simply referenced a single, but relevant, aspect of the story (e.g., only the protagonist’s mental state or only one situation related to the outcome) were given a score of 1. Responses based on irrelevant or incorrect information received a score of 0 (see supplementary material for full coding scheme). The total score was calculated separately for the eight mindreading stories and eight general comprehension stories, each with a maximum score of 16. Two researchers independently coded 30% of the participants’ responses to establish inter-rater reliability. Inter-rater reliability was excellent for both the mindreading stories, ICC = .92, 95% CI [0.92, 0.93], and general comprehension stories, ICC = .92, 95% CI [0.91, 0.93].
Accent manipulation check and evaluations
To check if participants correctly recognized the accents of the protagonists in the Strange Stories task as Australian-accented English or not, they were first asked to guess where the speakers were from. Participants were then asked, “How would you categorize the accent of the speakers in these stories?”. They responded on a 5-point scale (1 = definitely Australian, 5 = definitely not Australian). Participants in the Native Accent condition who responded “Somewhat/Definitely not Australian” and those in the Foreign Accent condition who responded “Somewhat/Definitely Australian” were excluded from analysis for failing the accent manipulation check. Participants who responded “Neutral” were only included if they identified the origin of the speakers accurately (Australia for Native Accent condition, Asia for Foreign Accent condition).
Participants were also asked to rate the nativeness, fluency, and comprehensibility of the protagonists in the Strange Stories task. For nativeness, participants were asked “How would you categorize the level of English of the speakers in these stories?” They responded on a 5-point scale (1 = definitely non-native, 5 = definitely native). For fluency, participants were asked “Please rate the fluency of the speakers in these stories.” They responded on a 5-point scale (1 = very disfluent, 5 = very fluent). For comprehensibility, participants were asked, “How easy was it to understand the speaker in these stories?” They responded on a 9-point scale (1 = extremely difficult to understand, 9 = extremely easy to understand).
Perceived Similarity task
Participants were presented with a series of audio recordings featuring the speakers from the Strange Stories task uttering, “Hello, how are you?” They were asked to “imagine the type of people who sound like this—not just these four people specifically—but other people who also sound like them more generally,” then completed eight questions on how similar they thought they were to these people. The questions were adapted from the Perceived Similarity scale (Zellmer-Bruhn et al., 2008) and assessed similarities based on aspects such as behaviors, personalities, and cultural/ethnic background. Participants provided their ratings on a 7-point scale (1 = totally different, 7 = totally similar). The perceived similarity score was calculated by averaging all item responses, with higher scores indicating that the participant rated the speaker as more similar to themselves.
Modern Racial Prejudice scale
We adapted Akrami et al.’s (2000) Modern Racial Prejudice scale to measure prejudicial sentiments against immigrants in Australia. The nine items assessed participants’ agreement with statements that represent the three primary ways through which prejudice covertly manifests in modern societies: denial of continuing discrimination (e.g., “Discrimination against immigrants is no longer a problem in Australia”), antagonism toward immigrants’ demands (e.g., “Immigrants are getting too demanding in the push for equal rights”), and resentment about special favors for immigrant populations (e.g., “Special programs are needed to create jobs for immigrants” [reverse coded]). Participants responded on a 7-point scale (1 = strongly disagree, 7 = strongly agree). Four items were reverse-coded, and the modern prejudice score was calculated as the average of all item scores. Higher scores indicated stronger prejudicial attitudes toward immigrants.
Assimilation and Integration scales
Nshom and Kahlimzoda’s (2020) Assimilation and Integration scales were adapted for the Australian context.
Assimilation
Four items assessed participants’ preference for immigrants to forgo their own culture and engage in Australian culture, exclusively with Australian people. Participants were asked questions such as, “Immigrants should adopt Australian culture and not keep their own”, and responded on a 7-point scale (1 = strongly disagree, 7 = strongly agree). The assimilation score was calculated by averaging all item responses, with higher scores indicating a stronger belief that immigrants should assimilate.
Integration
Four items measured participants’ preference for immigrants to maintain their own culture, while also engaging in Australian culture with other immigrants and Australian people. Participants were asked questions such as, “Immigrants should keep their own culture, but also adapt to Australian culture,” and responded on a 7-point scale (1 = strongly disagree, 7 = strongly agree). The integration score was calculated by averaging all item responses, with higher scores indicating a stronger belief that immigrants should integrate.
Accent Beliefs Scale
Hansen’s (2000) Accent Beliefs Scale was used to measure the degree to which participants believed an individual’s accent comprised a key, unchangeable component of that individual.
Accent stability
Four items measured participants’ beliefs about how permanent and stable an individual’s accent is over their lifetime. Participants identified their level of agreement with statements such as, “If someone was raised in one language, they will have its accent their whole life” on a 7-point scale (1 = strongly disagree, 7 = strongly agree). Two items were reverse-coded, and ratings from all items were averaged to yield the accent stability score. Higher scores indicated stronger beliefs that accent is stable across one’s lifetime.
Accent diagnosticity
Eight items assessed participants’ beliefs that they can infer an individual’s personal attributes (e.g., personality, intelligence) through the strength and type of accent the individual speaks with. Participants identified their level of agreement with statements such as, “It is possible to tell how someone will act by hearing their accent” on a 7-point scale (1 = strongly disagree, 7 = strongly agree). Four items were reverse-coded, and ratings from all items were averaged to yield the accent diagnosticity score. Higher scores indicated stronger beliefs that accent is diagnostic of an individual’s personal attributes.
Demographic and intergroup contact questionnaire
Participants were asked for basic demographic details (age, gender), including their racial-ethnic and linguistic background. Following information about themselves, they were asked a set of four questions about their contact with people from other backgrounds. In particular, as a measure of general exposure to people of other race and accent groups, participants were asked how often they come across people “. . . of a different race to you?” and “. . .who speak with a different accent (when speaking English) to you?”. Responses were provided on a 5-point scale (1 = never, 5 = always). Participants’ responses from these two items were averaged to derive an intergroup exposure score. As a measure of close contact they have with other race- and accent-group members, participants were asked how many of their friends “. . . are of a different race to you?” and “. . . speak with a different accent (when speaking English) to you?” Responses were given on a 5-point scale (1 = none, 5 = all). The mean rating from these two items represented the intergroup friendship score.
Procedure
Participants completed a Qualtrics survey which included the above measures. Participants were randomly allocated to either the Native Accent or Foreign Accent condition. All participants took part in the Strange Situations task first, followed by the accent manipulation check and evaluations questions, then the Perceived Similarity task. Participants were then presented with a block of questionnaires which included the Modern Racial Prejudice scale, Assimilation and Integration scales, and the Accent Beliefs Scale. The order of scales within the block was randomized. Lastly, participants completed questions on their demographic details and intergroup contact. Participation took approximately 45 minutes.
Attention check measures were integrated throughout the survey. In the Strange Situations task, four video clips ended with an animal sound (e.g., cow mooing) and participants were asked to identify the relevant animal. Participants who identified two or more of the animals incorrectly were excluded from analysis. In the block of questionnaires, one attention check was incorporated within each of the four scales, wherein participants were asked to select a specific response. Participants who identified any other response in any of the attention checks were excluded from analysis.
Results
Group Differences in Mindreading
To examine the effect of accent condition and participant accent on performance on the Strange Stories task, we conducted a 2 (Accent Condition: Native, Foreign) × 2 (Story Type: Mindreading, General Comprehension) × 2 (Participant Accent: Australian, Non-Australian) mixed ANCOVA (see Table 1 for descriptive statistics). We included participant age and gender as covariates, given the continued developmental changes in mindreading capacities across adulthood (e.g., Gourlay et al., 2024; Sebastian et al., 2012), as well as greater engagement of brain regions involved in perspective taking when completing mindreading tasks for women compared to men (e.g., Adenzato et al., 2017; Frank et al., 2015). This analysis revealed a significant main effect of accent condition, F(1, 550) = 6.29, p = .012, ηp2 = .01. Regardless of the participants’ own accent or story type, participants who heard native, Australian-accented protagonists scored higher (M = 13.43, SE = 0.13) than those who heard the foreign-accented protagonists (M = 12.98, SE = 0.12). There was also a significant main effect of Story Type, F(1,550) = 37.65, p < .001, ηp2 = .06, such that participants scored higher on mindreading stories (M = 13.57, SE = 0.09) compared to general comprehension stories (M = 12.85, SE = 0.12). Finally, the analysis revealed a significant main effect of Participant Accent, F(1, 550) = 8.41, p = .004, ηp2 = .02. Specifically, Australian-accented participants (M = 13.47, SE = 0.08) performed better on the Strange Stories task overall compared to participants who spoke with non-Australian accents (M = 12.95, SE = 0.16). There were no significant interactions (see Table S3 in supplementary material for statistical output). 2
Means and standard deviations for participants’ scores on the Strange Stories task, accent evaluations, Perceived Similarity task, prejudicial attitude measures, and intergroup contact as a function of accent condition and participant accent.
Notes. Strange Stories task (minimum = 0, maximum = 16); Nativeness (1 = definitely non-native, 5 = definitely native); Fluency (1 = very disfluent, 5 = very fluent); Comprehensibility (1 = extremely difficult to understand, 9 = extremely easy to understand); Perceived Similarity (1 = totally different, 7 = totally similar); Assimilation and Integration, Modern Racial Prejudice, Accent Beliefs Stability and Diagnosticity (1 = strongly disagree, 7 = strongly agree); Exposure to people of other race/accent (1 = never, 5 = always); Friendships with people of other race/accent (1 = none, 5 = all). For each measure, values marked with a were significantly greater than b.
See Tables S3 and S5 in supplementary material for statistical output.
Predictors of Mindreading
The descriptive statistics for participants’ evaluations of the protagonists’ accents, Perceived Similarity task, prejudicial attitude scores, and intergroup contact measures are presented in Table 1. A series of 2 (Accent Condition) × 2 (Participant Accent) ANOVAs were conducted on these measures. In support of the idea that the Malaysian- and Singaporean-accented protagonists were perceived to speak with characteristics typically associated with a “foreign” accent, participants in the Foreign Accent condition provided significantly lower ratings of nativeness, fluency, and comprehensibility compared to those in the Native Accent condition. The main effect of accent condition on these ratings was found, regardless of whether the participants themselves spoke with an Australian accent or non-Australian accent. Furthermore, no main effects of accent condition or participant accent were found for most intergroup factors—this included participants’ scores on the Perceived Similarity task, Modern Racial Prejudice scale, Assimilation scale, Accent Beliefs Diagnosticity subscale, and Intergroup Exposure (see Table S5 in supplementary material for statistical output).
To investigate how these measures linked to participants’ mindreading performance, we first ran bivariate correlations between all predictor variables and participants’ scores on the mindreading stories in the Strange Stories task. As can be seen in Table 2, mindreading was significantly correlated with five variables: participants’ fluency rating of the protagonists in the Strange Stories task, scores on the Modern Racial Prejudice scale, Assimilation scale, Accent Beliefs Diagnosticity subscale, and reported amount of intergroup friendships.
Bivariate correlations between all predictor variables and mindreading.
Notes. *p < .05, **p < .01.
Next, we conducted a hierarchical multiple regression analysis to examine if any variable identified to be a significant correlate uniquely predicted mindreading. All continuous variables were mean-centered for analysis. At Step 1, we entered accent condition (Native, Foreign) and participant accent (Australian, non-Australian), as well as participant demographic variables (age, gender) as covariates. To control for participants’ baseline capacities for understanding the speakers and following narratives to reason accordingly, we also included participants’ scores on the general comprehension stories of the Strange Stories task as a covariate. At Step 2, we entered the five predictor variables which were identified as significant correlates of mindreading by the bivariate correlations—namely, participants’ fluency rating, scores on the Modern Racial Prejudice scale, Assimilation scale, Accent Beliefs Diagnosticity subscale, and intergroup friendship. Finally, given that we expected these predictor variables to be linked to mindreading when taking the perspective of foreign-accented speakers in particular (and not native-accented speakers), we entered interaction variables that crossed each of our predictor variables with accent condition in Step 3.
At Step 1, participant age, gender, accent condition, participant accent, and general comprehension scores together explained a significant 21.90% of variance in mindreading, F(5, 549) = 30.83, p < .001. As can be seen in Table 3, participants’ score on the general comprehension stories of the Strange Stories task was found to be a significant predictor of mindreading, whereby those who scored higher on general comprehension stories also scored higher on mindreading stories. Participant age, gender, accent condition, and participant accent were not significant predictors of mindreading.
Hierarchical multiple regression analysis for variables predicting mindreading in the Strange Stories task.
At Step 2, the model additionally including fluency, prejudice, assimilation, accent diagnosticity, and intergroup friendship together explained a significant 24.30% of variance in mindreading, F(10, 544) = 17.42, p < .001. The additional 2.30% of variance significantly increased the explanatory power of the model, F ch.(5, 544) = 3.35, p = .005. As can be seen in Table 3, participants’ scores on the general comprehension stories of the Strange Stories task remained a significant predictor of mindreading. Additionally, participants who reported to have more intergroup friendships performed worse on the mindreading stories of the Strange Stories task. All remaining variables were non-significant predictors of mindreading.
At Step 3, the model including the interaction variables together explained a significant 25.90% of variance in mindreading, F(15, 539) = 12.54, p < .001. The additional 1.60% of variance significantly increased the explanatory power of the model, F ch.(5, 539) = 2.35, p = .040. As can be seen in Table 3, once again, participants’ scores on the general comprehension stories of the Strange Stories task and amount of intergroup friendships were significant predictors of mindreading. Additionally, the Accent Condition × Intergroup Friendship interaction was found to be a significant predictor. Holm-Bonferroni corrections were applied to the family of tests related to intergroup friendships (Bender & Lange, 2001); the significance of these predictors remained. The remaining variables were non-significant.
To follow up the significant interaction in the overall hierarchical regression analysis, we conducted two separate regressions: one for the Native Accent condition and one for the Foreign Accent condition. Consistent with the overall regression analysis, participant demographic variables (age, gender, accent) and general comprehension scores were entered at Step 1. The five predictor variables—fluency, prejudice, assimilation, accent diagnosticity, and intergroup friendship—were entered at Step 2. The full regression output for both analyses are shown in Table 4.
Hierarchical multiple regression analyses for variables predicting mindreading of native-accented and foreign-accented protagonists.
Native Accent condition
At Step 1, participant age, gender, accent, and general comprehension scores significantly explained 18.20% of variance in mindreading, F(4, 263) = 14.62, p < .001. Participants who scored higher on the general comprehension stories also scored higher on mindreading stories on the Strange Stories task. All other variables entered at Step 1 were non-significant predictors.
At Step 2, the model including fluency, prejudice, assimilation, accent diagnosticity, and intergroup friendship together explained a significant 19.50% of variance in mindreading, F(9, 258) = 6.95, p < .001. The additional 1.30% of variance explained at Step 2, however, was not significant, F ch.(5, 258) = 0.86, p = .512. Participants’ scores on the general comprehension stories remained a significant predictor of mindreading. None of the other variables were significant predictors.
Foreign Accent condition
At Step 1, participant age, gender, accent, and general comprehension scores together significantly explained 24.80% of variance in mindreading, F(4, 282) = 23.30, p < .001. Consistent with findings on the Native Accent condition, participants’ performance on the general comprehension stories of the Strange Stories task was again a significant, positive predictor of mindreading foreign-accented protagonists. All other variables entered at Step 1 were non-significant predictors.
At Step 2, the model including fluency, prejudice, assimilation, accent diagnosticity, and intergroup friendship together explained a significant 30.5% of variance in mindreading, F(9, 277) = 13.50, p < .001. The additional 5.60% of variance explained at Step 2 was significant, F ch.(5, 277) = 4.50, p = .001. Participants’ scores on the general comprehension stories remained a significant predictor of mindreading. In contrast to the findings on the Native Accent condition, for participants who listened to foreign-accented protagonists, those who reported to have more intergroup friendships were worse at taking the perspectives of the foreign-accented protagonists in the Strange Stories task.
Discussion
In the present study, we examined the impact of accent-based group membership on mindreading in adults. Based on findings from previous studies which suggest we are better at taking the perspective of people from our own cultural group than of those from other groups (Ekerim et al., 2020; Gönültaş et al., 2019; Perez-Zapata et al., 2016), we expected that Australian adults would perform better at identifying the mental states of native, Australian-accented protagonists compared to foreign, Malaysian- and Singaporean-accented protagonists in the Strange Stories task. Our findings support this hypothesis and provide the first empirical evidence of the effect of accent on mindreading. Even though previous studies did not find that accent-based cultural group membership impacts mindreading in children (Anderson et al., 2023; Witt et al., 2022), our findings suggest that accent influences how effectively we take the perspectives of others in adult interactions.
Contrary to previous studies which found the effect of cultural group membership on mindreading specifically (Ekerim et al., 2020; Gönültaş et al., 2019; Perez-Zapata et al., 2016), participants who listened to native-accented protagonists outperformed those who heard foreign-accented protagonists not only on stories that required perspective taking but also on general comprehension stories void of mental states. One explanation for this may be offered by Construal Level Theory (Trope & Liberman, 2010). This theory suggests that we tend to represent situations involving people who are socially distant (e.g., those from different cultural groups) in abstract, decontextualized terms. As a result, we resort to dispositional attributions about the behaviors of socially distant others and chunk sequences of events involving these individuals into broad, unsegmented representations (for a review, see Trope & Liberman, 2010). This may explain why participants who listened to foreign-accented protagonists did not integrate as much relevant situational information in their responses compared to those who heard the native-accented protagonists, regardless of whether the information pertained to mental states or not.
Alternatively, the finding that participants performed better overall on the Strange Stories task when listening to the native- over foreign-accented protagonists raises the possibility that participants simply could not understand the stories as well when narrated with foreign accents. There are several reasons why we believe this to be unlikely. First, all auditory stimuli used in this study were selected based on high levels of comprehensibility (ease of understanding) and intelligibility (can be accurately transcribed verbatim) identified by pre-testing the stimuli. Second, comprehensibility ratings provided by the participants in this study reiterate this preliminary finding: although mean ratings were slightly higher for native- compared to foreign-accented protagonists, participants rated both to be “very easy to understand” (i.e., average of around 8 out of 9). Third, even in cases where participants may have had difficulties processing the auditory narrations, subtitles were visually provided in every story in both accent conditions to scaffold participants’ comprehension (Burchill et al., 2018). For these reasons, the main effect of accent found in the Strange Stories task is likely not a simple product of participants being unable to effectively understand the stories narrated by foreign-accented protagonists. Nevertheless, we subsequently controlled for participants’ general comprehension demonstrated in the Strange Stories task when investigating the links between mindreading and the individual predictors.
In addition to the accent of the protagonists in the Strange Stories task, the incidental recruitment of participants who spoke English with non-Australian accents allowed us to conduct a preliminary exploration of how participants’ own accent may impact their performance on the task. Overall, participants who spoke English with an Australian accent performed better on the Strange Stories task than those who spoke with other accents. The most parsimonious explanation is that the majority of non-Australian-accented participants were non-native English speakers; thus, on average, they may have been less linguistically proficient than the Australian-accented participants in listening to the stories and providing the required responses. Moreover, regardless of whether the participants themselves spoke English with an Australian accent or not, we found that they performed better when hearing the stories told by the native, Australian-accented protagonist over the foreign-accented counterparts. On one hand, this finding does not support the idea that the effect of cultural group membership on mindreading is driven by shared group membership (Perez-Zapata et al., 2016), at least in terms of overlap in accent. On the other hand, all our participants live in Australia, so even our non-Australian-accented participants may have felt a closer social distance to the native, Australian-accented protagonists than the foreign-accented protagonists. This may have been sufficient to activate the social-cognitive tendencies explained by Construal Level Theory (Trope & Liberman, 2010). Given the relatively small sample of non-Australian-accented participants in our study, future research must be conducted to more comprehensively examine how listener and speaker accents interact to determine social outcomes.
What Underlies the Accent Effect on Mindreading?
When taking the perspectives of native-accented protagonists, we found that participants’ general comprehension ability was the only significant predictor of their mindreading in the Strange Stories task. In other words, intergroup factors did not impact participants’ accuracy in reading the minds of native-accented protagonists. In contrast, even when controlling for variations in general comprehension ability, participants’ mindreading of foreign-accented protagonists was related to a measure of intergroup contact; specifically, the number of friends they reported to have who are of other races or speak English with other accents to their own. The directionality of this relation, however, was the opposite of what we had expected: we found that participants who reported to have more intergroup friendships performed worse at identifying the mental states of foreign-accented protagonists.
Our finding on the negative association between intergroup friendship and mindreading of foreign-accented individuals stands in contrast to our original prediction based on the intergroup contact theory (Pettigrew, 1998). This theory suggests that contact with people of other social groups—especially when experienced in the form of close social relationships (e.g., friendships; Davies et al., 2011)—reduces prejudicial attitudes toward those groups. This, in turn, should encourage efforts to take the perspectives of people belonging to those groups (Ekerim et al., 2020; Gönültaş et al., 2019). In our study, we found that the reported amount of exposure in general to people of other races and accents was negatively linked to prejudice. Close intergroup contact via friendships, however, was not related to any of our prejudice-related measures. A different mechanism to what we had originally expected, therefore, may be at play in explaining the role of intergroup friendships on mindreading.
Construal Level Theory may, once again, afford an explanation for why participants who had more intergroup friends performed worse on mindreading the foreign-accented protagonists. According to this theory, when psychological distance is perceived in situations that are otherwise expected to feel proximal, this mismatch can amplify the sense of distance (Trope & Liberman, 2010). In the context of our study, pre-existing social closeness via friendships with people of different races and accents may have conflicted with the social distance elicited by the foreign-accented protagonists in the Strange Stories task. This, in turn, may have increased participants’ tendencies to abstractly represent the situations involving the foreign-accented protagonists, leading to poorer task performance.
An alternative framework for interpreting our finding may be provided by the growing research on the role of effort in mindreading (Anderson et al., 2023; Hawkins et al., 2021; Savitsky et al., 2011; Simpson & Todd, 2017). Perspective taking is a cognitively effortful task; this was likely accentuated in our Strange Stories task, since listening to foreign-accented speech itself is often perceived to be effortful (Peng & Wang, 2019). Given the mind’s propensity to conserve cognitive efforts where possible (Gheza et al., 2023; Zekveld & Kramer, 2014), we tend to relax our perspective taking to reduce cognitive costs when we feel that we have a privileged understanding of our communicative partner (Hawkins et al., 2021). Following this line of thought, our participants who had more other-race and other-accented friends may have perceived the foreign-accented protagonists in the Strange Stories task as people they can readily understand. In turn, they may have minimized their efforts to mindread these individuals.
Limitations and Future Directions
Our study provides the first evidence that speaks to the role that accents may play in understanding each other when interacting with people of diverse cultural backgrounds. In particular, our findings were based on Malaysian- and Singaporean-accented English representing foreign accents within the Australian context. These parameters should be kept in mind when interpreting the findings, given that different accents carry varying social meanings within the contexts in which they are encountered (Dragojevic & Goatley-Soan, 2020). For instance, in contrast to previous studies that have found the effect of cultural group membership on mindreading specifically (Ekerim et al., 2020; Gönültaş et al., 2019; Perez-Zapata et al., 2016), we found that accent also influenced participants’ performance on the general comprehension stories on the Strange Stories task. On one hand, our more generalized findings may reflect the broader impact of accent as a linguistic barrier to social interactions. On the other hand, previous studies suggest that intercultural mindreading bias was driven primarily by prejudicial attitudes (Ekerim et al., 2020; Gönültaş et al., 2019)—links that we did not observe in our study, likely due to the overall neutral responses participants provided on our prejudice-related measures. Given the large number of immigrants from Malaysian and Singapore (Australian Bureau of Statistics, 2021) as well as the social norms against racism in Australia (Ben et al., 2024), the foreign, Malaysian- and Singaporean-accented protagonists in our study may not have elicited the prejudicial sentiments necessary to observe the bias on mindreading specifically. Future studies involving various combinations of accent varieties, participant background, and social-cultural contexts are needed to develop our understanding of the conditions under which accent impacts perspective taking.
Relatedly, future studies could investigate the nuances in prejudice-related attitudes that predict intergroup mindreading. In our study, we opted for indirect measures of prejudice (e.g., “Discrimination against immigrants is no longer a problem in Australia”) because overt racism is frowned upon in Australia (Ben et al., 2024). In contrast, previous studies which found prejudice to be a predictor of intercultural mindreading measured explicitly negative attitudes against the target cultural group that are arguably dehumanizing in nature (e.g., “Syrian immigrants in Turkey do not take care of their personal hygiene”; Ekerim et al., 2020). Dehumanization undermines attributions of mental states (Harris & Delgado, 2025), so it stands to reason that measures that tap into these sentiments may be especially predictive of poor intercultural mindreading. In modern societies, the manifestation of prejudicial attitudes is complex and context-specific (Akrami et al., 2000). Exploring how mindreading is linked to these various expressions, therefore, will pave the way to a more comprehensive understanding of how intergroup attitudes influence cross-cultural interactions.
Finally, while previous studies have focused on how “othering” hinders mindreading, our study explored for the first time how intergroup contact—theorized to bridge people of different social groups—impacts perspective taking. Although the outcome of reducing social distance via intergroup contact has predominantly been studied in relation to prejudice mitigation (Pettigrew et al., 2011), our study provides preliminary evidence suggesting that social closeness with people of different races and accents may have other effects on cross-cultural interactions. Namely, we interpreted the negative association between intergroup friendship and mindreading in light of the literature that emphasizes the role of effort in successful perspective taking (Anderson et al., 2023; Hawkins et al., 2021; Simpson & Todd, 2017). Our interpretations are largely speculative, however, given that we did not directly measure the effort that participants put into the Strange Stories task. Through a combination of methodological approaches including self-report and physiological measures (McGarrigle et al., 2014), future research could investigate if and how effort is modulated when taking the perspectives of people from different backgrounds. Furthermore, for mindreading situations involving foreign-accented speech, it would be important to delineate between effort invested in listening to the speech versus the higher-order process of mindreading. This delineation will allow researchers to explore the possibility that communication challenges in intercultural settings may at least in part be a product of listeners simply “tuning out” when hearing foreign-accented speech (Lee, 2023).
Conclusion
When we come across people of diverse cultures in our everyday social interactions, we often make spontaneous inferences about their background, simply by the way they speak. As adults, we not only use accents to identify people’s geographical origin and cultural preferences, but these inferences prompt stereotypes and societal attitudes that shape the way we interact (Fuertes et al., 2012; Gluszek & Dovidio, 2010). The findings from our study suggest that one such way in which accent bias manifests in our social interactions is via creating psychological distance between people who speak with different accents; this, in turn, may impact how closely we listen to and put in the effort to understand them.
Supplemental Material
sj-docx-1-gpi-10.1177_13684302261426266 – Supplemental material for Accent bias in mindreading
Supplemental material, sj-docx-1-gpi-10.1177_13684302261426266 for Accent bias in mindreading by Christopher B. Buckland, Jessica L. Spence and Kana Imuta in Group Processes & Intergroup Relations
Supplemental Material
sj-xlsx-2-gpi-10.1177_13684302261426266 – Supplemental material for Accent bias in mindreading
Supplemental material, sj-xlsx-2-gpi-10.1177_13684302261426266 for Accent bias in mindreading by Christopher B. Buckland, Jessica L. Spence and Kana Imuta in Group Processes & Intergroup Relations
Footnotes
Acknowledgements
We would like to thank Evena Wong for the illustrations featured in our stimuli, Jin Yi Loh and Jizelle Ellul for assistance with integrating the audio and visual components of the stimuli, Joanna Sams for her help with coding the data, and Dr. Brendan Zietsch for statistical consultation.
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.
Ethical approval and informed consent
This study was approved by The University of Queensland Human Research Ethics Committee (Approval number: 2020/HE001178). All participants completed an online written consent form prior to taking part in the study.
Data availability statement
Data set is available in the Supplementary Data file.
Supplemental material
Supplemental material for this article is available online.
Notes
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
