ABSTRACT

Studies have found that bilinguals respond differently to personality measures in their two languages, indicating that bilinguals change their personality as they switch between their two languages and/or cultures. Across two experiments, we attempted to investigate the effect of language and culture separately on how bilingual speakers rate themselves on the personality dimensions on the Big-5 Personality Inventory. Swedish speakers were asked to imagine applying for a job either at a Swedish (home country/own culture) or an American (foreign country/culture) company, and they responded to the Big-5 questions in either their first language Swedish, or their second language English, in a 2 by 2 design. Overall, differences on several of the personality dimensions were found, mainly affected by the language factor, generally replicating previous research. These results suggest that separate processes may be driving previously found differences on personality measures in bilinguals’ two languages to some extent, and that these processes affect the personality dimensions (as measured by the Big-5 personality inventory) differently, even if the language variable seems to be the stronger indicator.

Introduction

As bilingualism has become increasingly common world-wide, a steady stream of studies has researched the duality associated with bilinguals’ different languages in relation to a number of not only linguistic, but also non-linguistic, aspects. For example, effects of language have been found in terms of colour perception and categorisation (e.g. Athanasopoulos et al. Citation2010; Roberson et al. Citation2005; Winawer et al. Citation2007), time perception (e.g. Bylund and Athanasopoulos Citation2017; Miles et al. Citation2011), motion encoding (e.g. Athanasopoulos et al. Citation2015), level of experienced emotionality (e.g. Caldwell-Harris Citation2014; Dewaele Citation2010; Dylman & Bjärtå Citation2019; Pavlenko Citation2005), decision-making processes (e.g. Costa et al. Citation2014; Keysar et al. Citation2012), and responses to moral dilemmas and other problem-solving scenarios (e.g. Corey et al. Citation2017; Costa et al. Citation2014; Dylman & Champoux-Larsson Citation2020).

Another domain that has been investigated in relation to differences between bilinguals’ two languages concerns personality, with the pertinent question of whether, as Ramirez-Ezparza et al. (Citation2006) expressed it, bilinguals have different personalities in their two languages. More specifically, Ramírez-Esparza and colleagues examined cultural frame switching in Spanish-English bilinguals in North America. Cultural frame switching refers to the phenomenon whereby ‘bicultural individuals shift values and attributions in the presence of culture-relevant stimuli’ (Ramirez-Ezparza et al. Citation2006, 100), following research showing that cultural cues can make salient specific attributes and values associated with said culture (e.g. Benet-Martinez & John Citation1998; Hong et al. Citation2000). In their study, Ramírez-Esparza et al. measured personality traits using the Big-5 personality inventory, and employed a within-subjects design where Spanish-English bilinguals in Mexico and the US completed the Big-5 personality questionnaire in both Spanish and English. Across three experiments with independent samples of Spanish-English bilinguals, Ramírez-Esparza et al. found cross-language differences for most dimensions, all of which consistent with monolingual samples in each respective culture, i.e. Americans or answering in English scored higher on extraversion, agreeableness, conscientiousness, and lower on neuroticism than Mexicans or answering in Spanish. These results suggest that the language context can lead to a cultural frame shift in bilinguals whereby they attune their attitudes and values to correspond with the culture of a given language context. Similar effects of language on aspects of personality and self-perception have been reported by Chen et al. (Citation2014) who asked Chinese-English bilinguals in Hong Kong to complete measures of self-efficacy, and the conscientiousness subscale from the Big-5 questionnaire, either in Chinese or in English. The bilinguals who completed the questionnaires in English reported higher levels of both self-efficacy and conscientiousness compared to the bilinguals who completed the measurements in Chinese, again consistent with the cultural meaning systems of each respective culture (e.g. Chen et al. Citation2014; Lee et al.Citation2010). Other studies have also highlighted the notion of the bilingual identity being affected by language mode and/or cultural context. For example, Panayiotou (Citation2004) presented a scenario to bilinguals in either English or Greek and either in an American or a Greek setting. The participants were then interviewed and asked about their emotional responses to the story. The results found that the participants’ emotional responses changed depending on the language (and culture) in which the story was set. Likewise, Ramírez-Esparza et al. (Citation2008) investigated simpatía, a cultural script in the Hispanic culture, which can be described as being similar to the Big-5 dimension agreeableness. The paradoxical question they investigated was why Hispanics score lower on this dimension despite often being characterised as simpatíco (agreeable). They found that Mexican-English bilinguals displayed more agreeable behaviours in Spanish than in English, and yet self-reported themselves as being lower in agreeableness when assessed in Spanish than in English. The results were explained in terms of modesty being an important part of sympatía, which in turn lead to the participants scoring themselves as lower in sympatía in Spanish despite behaving more agreeable. These results demonstrate the complexity of the interaction between culture, language, and behaviour.

According to the lexical paradigm in personality psychology (Goldberg Citation1981) it is assumed that individual differences within a cultural context are encoded in everyday language, as such differences need to be communicated within the culture. Accordingly, similar concepts may have different connotations across cultures (Kitayama & Park Citation2007). The Big-5 personality theory, mentioned above, seems to converge universally regarding four of the five dimensions, while the dimension most often labelled Openness has been more difficult to reproduce within different cultural contexts (Heine & Buchtel Citation2009; Hendriks et al. Citation2003). Originally developed within a Western context, and measured with instruments translated from English (Heine & Buchtel Citation2009), it is evident that differences in levels of the five dimensions could reflect a number of possibilities. They could be reflections of real cultural differences, different interpretations of the items, differences in response patterns, or comparisons with different standards (Dylman et al. Citation2020). It is thus unclear whether the above-mentioned effects of language on personality traits are the result of a change in language or culture, or both, and if so, what the individual contribution of language and culture is on observed effects. Indeed, some researchers have argued that the typically observed higher levels of emotional reactivity to emotional stimuli (such as taboo words) in a first language (L1) compared to a second language (LX) is due to cultural and social norms, rather than language per se (e.g. Gawinkowska, Paradowski, and Bilewicz Citation2013). Even if Panayiotou’s (Citation2004) study specifically mentioned both language and culture, and to some extent separated the two, they were largely overlapping and the design of that study did not allow for a systematic investigation of the different combinations of language and culture in a two by two design. Thus, trying to disentangle culture and language may be informative, both methodologically and theoretically, to explain various phenomena relating to observed differences associated with bilinguals’ two languages.

Cultural differences have been found regarding the Big-5 in numerous studies (for a review, see Heine & Buchtel Citation2009). Personality profiles from cultures from similar geographical and historical origin seem to cluster together (Allik and McCrae Citation2004), but not all similarities can be explained by such dimensions (Schmitt et al. Citation2007). A recent study (Allik et al. Citation2017) reanalyzing data from 62 countries in 37 different languages, found clear groupings of countries, but also found that between-country individual differences were smaller than within-country individual differences. Some research has also tried to correlate national means of the Big-5 to measures of cultural differences, for example the four dimensions proposed by Hofstede and colleagues (Minkov & Hofstede Citation2011), namely, power distance, individualism-collectivism, masculinity-femininity, and uncertainty avoidance. These have been found to correlate with almost all Big-5 dimensions (Bartram Citation2013). Especially the variability of personality scores shows consistent patterns – the higher the power distance (i.e. social inequality, relations with authority) within a country, the lower the variability in personality, and the higher the level of individualism the higher the degree of variability in personality. According to these dimensions, the United States and Sweden place themselves in the same cultural space (Allik et al. Citation2017), although Americans are found to be higher on individualism, power distance, uncertainty avoidance and masculinity (Minkov & Hofstede Citation2011). The biggest difference is for masculinity on which Sweden scores extremely low (Hornsey et al. Citation2019). Other cultural differences found between Sweden and the US is that Swedes are found to have higher levels of interpersonal trust (Allik & Realo Citation2004), and to express lower levels of stress, depression and anxiety (Scholten et al. Citation2018). Furthermore, differences between Europeans and Americans have been found regarding their attitudes towards work (Okulicz-Kozaryn Citation2011). According to this, while Sweden and the US are, from a global perspective, more similar than different, some cultural differences are to be expected. For example, the huge difference in the dimension of masculinity would suggest differences between Americans and Swedes similar to those found between men and women, where women tend to score higher than men on Agreeableness, Extraversion and Neuroticism (e.g. Costa et al. Citation2001; Lippa Citation2010; Schmitt et al. Citation2008; Weisberg et al. Citation2011). Significant negative correlations have indeed been found between the cultural dimension of masculinity and emotional stability, extraversion and agreeableness (Bartram Citation2013).

In the current study, we attempted to examine the notion of bilinguals having different personalities in their different languages more closely by systematically distinguishing between the unique role of language and culture separately. On a more general level, we also attempted to replicate the previous findings of differences in personality in bilinguals’ different languages in another population, namely Swedish speakers with English as a second language. In Experiment 1, we designed an experiment where L1 Swedish speakers were asked to imagine going for a job interview. The scenario was presented either in the participants’ L1 (Swedish) or in their LX (English), and the imagined job interview was either at a Swedish company or at an American company, in a 2 by 2 design. Following the scenario, the participants were asked to fill out the Big-5 personality inventory in the same language as the scenario they had just read. This allowed us to investigate the effect of language and culture separately. If a language effect exists, we would expect a main effect of language, observed as differences between L1 and LX regardless of cultural context (Swedish or American). If a cultural effect exists, we would expect a main effect of culture, observed as differences between the two culture manipulations (Swedish vs. American) regardless of the language (Swedish or English) in which the questionnaire is answered.

Experiment 2 attempted to replicate Experiment 1, using predominantly the same methodology. To this end, a different group of L1 Swedish speakers with English as an LX were asked to imagine going for a job interview and to fill in the Big-5 personality inventory. The main difference between Experiments 1 and 2 was that the language and culture manipulations were made more salient in Experiment 2, and this was done in two ways. First, we included three generic open-ended interview questions to be answered. These questions were presented prior to the Big-5 personality inventory. Second, the flag of the country (and thereby culture) of the company was displayed in the top right corner during the presentation of the cultural manipulation (the scenario presenting the imagined company and the three interview questions).

Although the current study must be regarded as exploratory, some expectations may be outlined. If there is a main effect of culture, differences between company contexts (Swedish vs. American) would be expected, regardless of in which language the task is presented. As the cultural gap is widest for the cultural dimension of masculinity between Sweden and the US (Hornsey et al. Citation2019), personality differences would be expected for the same dimensions that differ between men and women, i.e. agreeableness, neuroticism, and openness (Bartram Citation2013). A main effect of language, however, would manifest itself as differences between the first- (Swedish) and second language (English) condition, regardless of the cultural context. As the first language seems to be more connected to emotional processes (e.g. Caldwell-Harris Citation2014; Dewaele Citation2010; Dylman & Bjärtå Citation2019; Pavlenko Citation2005; Panayiotou Citation2004), differences would be expected primarily for neuroticism, but perhaps also for agreeableness, and to some extent extraversion.

Experiment 1: Swedish-English bilinguals

As mentioned in the Introduction, Experiments 1 aimed to further investigate previously found effects of cultural frame switching in bilinguals (see e.g. Chen et al. Citation2014; Ramirez-Ezparza et al. Citation2006) by attempting to separate language and culture. These previous studies have employed a methodological design whereby bilingual participants have filled out personality tests in one of their two languages, but as mentioned, this hinders insight into whether the observed cultural frame switch is due to a switch in language or culture, as these two strongly overlap.

Thus, Experiment 1 attempted to separate the two variables language and culture by presenting scenarios asking the participants to imagine applying for a job, either at a Swedish company or at an American company, and the scenario was either presented in Swedish (the participants’ L1), or in English (the participants’ LX). As part of this imagined job application process, the participants were asked to fill out the Big-5 personality inventory, presented in the same language as the language manipulation.

Experiment 2: Swedish-English bilinguals (strengthened saliency)

Experiment 2 was an attempt to further strengthen the saliency of the manipulation. Firstly, three open-ended questions were added to be answered prior to the Big 5 personality inventory, in order to make the imagined job recruitment scenario more salient. In addition, a coloured flag (either the Swedish or the US flag) was added to the top right side of the screen during the presentation of the scenario, the three open ended questions, and the Big 5 personality items, as an attempt to further strengthen the saliency of the company culture.

General discussion

The current study investigated the effects of language and culture on bilinguals’ responses to the Big 5 personality inventory. Previous studies have found that bilinguals respond differently to personality tests depending on the language in which they fill out the personality tests (e.g. Chen et al. Citation2014; Ramirez-Ezparza et al. Citation2006). Given the design of those studies, however, it has been difficult to disentangle the separate effects of language and culture. Thus, the main aim of the current study was to experimentally separate the effect of language and culture in a 2 by 2 design, namely; Swedish speakers were primed with either a Swedish or an American culture, and filled out the Big 5 personality questionnaire in either Swedish (their L1) or English (their LX). Across two experiments, we found effects of language and culture on various of the dimensions on the Big 5 personality inventory. In Experiment 1, participants rated themselves as more extraverted when filling out the personality questionnaire in English compared to Swedish. They also rated higher on neuroticism when filling out the personality questionnaire when imagining applying for a job at a Swedish compared to an American company, regardless of the language in which they filled out the questionnaire. In Experiment 2, using a somewhat different methodology where the priming manipulations were strengthened, participants rated themselves as more agreeable and more conscientious when filling out the personality questionnaire in Swedish, their L1. In combination, these effects generally replicate previous findings of cultural frame switching, suggesting that bilinguals’ responses to personality questionnaires are affected by the language in which they complete the questionnaire (Chen et al. Citation2014; Ramirez-Ezparza et al. Citation2006). In line with previous research, we found no effect on the dimension Openness. The effects found are to some extent in line with our exploratory expectations, where we expected, and found, differences on the dimensions neuroticism for the cultural manipulation, and differences on the dimensions agreeableness and extraversion for the language manipulation.

Our results also indicate that there may be separate processes driving previously reported findings of cultural frame switching (at least in some populations of bilinguals), and that language and culture may need to be more closely investigated and controlled for, before more comprehensive conclusions can be drawn in terms of which dimensions of the Big 5 personality inventory are affected by language and by culture respectively. It is worth pointing out here that the term bicultural does not necessarily mean bilingual, even if ‘bilinguals tend to be bicultural (e.g. LaFromboise et al. Citation1993)’. A person can be bicultural (having and switching between two different cultures) whilst only speaking one language. Likewise, a person can be bilingual without necessarily being bicultural, even if these instances are not the norm. Thus, different combinations in terms of biculturalism and bilingualism, as well as specific combinations of cultures and/or languages may be affected to differing degrees. The current study offers insight into one such population.

Relatedly, an important point to make here concerns precisely the specific population investigated in this study. Both experiments reported in this paper investigated Swedish-English bilinguals, or more specifically, Swedish speakers with English as their LX, as taught in Sweden. However, it is potentially important to note that this group are highly proficient in English as a foreign language, given findings that language proficiency, experience and immersion may modulate emotional reactivity (e.g. Caldwell-Harris Citation2014; Čavar & Tytus Citation2018; Degner et al. Citation2012; but also see Circi et al. Citation2021 and Del Maschio et al. Citation2022), for two meta-analyses both of whom failed to find a modulating effect of LX proficiency on the foreign language effect in decision making. Furthermore, previous research on this population (e.g. Dylman & Champoux-Larsson Citation2020) has contended that the cultural influence of English in Swedish society (in terms of music, tv-shows, social media etc.) has notable effects on the relationship between language and emotion. Dylman & Champoux-Larsson (Citation2020) reported several experiments investigating the foreign language effect in decision making (both in terms of framing effects in problem solving, and for moral dilemmas) in different samples from the same population (i.e. L1 Swedish speakers with LX English). Across several experiments, the authors report a failure to find a foreign language effect in this population using this specific language combination (Swedish-English). In other words, the participants displayed comparable emotional reactivity to their LX English as they did to their L1 Swedish, indicating that there is something special about the role of English in Sweden. This is important considering the explanation commonly given to explain the foreign language effect; namely that an L1 is acquired in emotionally varied situations leading to the co-development of emotion and language, whilst an LX becomes less embodied as it as it is typically acquired in a neutral classroom setting (e.g. Ivaz et al. Citation2015). Given that Swedes are frequently exposed to English via media typically aimed at evoking emotions, English (despite technically being an LX) may become more closely intertwined with emotional processes. Having said that, other studies on emotion processes in a Swedish population with LX English have indeed found differences between Swedish and English (e.g. Dylman & Bjärtå Citation2019; Dylman & Champoux-Larsson Citation2020), indicating that specific aspects of emotion processing (e.g. decision making vs. experienced distress during reading of negative texts) may tap into different mechanisms, and thus, yield varying observable results. Nevertheless, the Swedish and the American cultures are distinct, and in the current study we did, indeed, observe an effect of culture on the dimension of neuroticism in Experiment 1. In Experiment 2, this effect was likely cancelled out due to the increased saliency of the language manipulation, which may have decreased the priming of the culture manipulation.

The differences (between Experiment 1 and 2) in terms of which dimensions were affected by the experimental manipulation is, as mentioned, likely due to the differences in methodology, with the added flags on the screen in Experiment 2 and where the participants had the additional task of answering three open-ended job interview style questions. While the initial idea was that these open-ended questions would strengthen the saliency of the culture priming (by being made more aware of the scenario of being in a setting of applying for a job at a large Swedish or American company), instead, it likely had the effect of strengthening the language manipulation, as it involved using language to a larger extent. In fact, the questions largely revolved around reflecting on and describing desirable personal attributes in an imagined job interview setting, such as ‘What are your personal strengths?’ and ‘How do you think others would describe you?’. Even the questions about self-improvements in this context (‘Which personal characteristics would you like to develop and improve?’) is arguably priming the participants towards imagining future positive attributes. Thus, this may have primed the participants to think of themselves as having more agreeable and conscientious attributes, leading them to rate themselves higher on these two dimensions in an L1 language setting. Previous research has indeed found that using an LX to process a text decreases emotionality (e.g. Dylman & Bjärtå Citation2019). While the current study did not investigate emotionality as such (nor study negative emotions as in the Dylman & Bjärtå study), it is probable that a job interview setting involves emotional arousal, and that attending to positive and desired attributes involves similar mechanisms in terms of the effect of language. Future research should more closely investigate this idea by including a form of emotion measure as a complement when investigating the effect of language and culture on personality dimensions.

Another potentially crucial aspect to mention concerns the manipulation of the culture variable. In the current study, we framed this in terms of the working culture of a company (working under the assumption that the Swedish and the US American work culture are perceived as being distinct from each other). However, there may be additional ways of experimentally manipulating the culture, and thereby separating the variables language and culture. For example, future studies could look at geographical setting by comparing samples living in two different countries (and thus being immersed in different cultures). It would be particularly interesting to investigate cases where one of the variables (e.g. culture) differs but the other one (e.g. language) is held constant, such as comparing, for instance, the UK and the US (or Portugal and Brazil, and so on), or comparing, for example, Swedish and American companies in Sweden (or the US). Furthermore, the cultural setting can be made more salient by using experimental methods such as presenting audio-visual cues of, for example, Swedish or American work settings. Whether these manipulations would change the results remains an empirical question for now. Furthermore, in the current study, we chose American (rather than British) as the foreign culture for two reasons: (1) The UK being in Europe could be perceived as being closer to the own culture, and given the exploratory nature of this study, we wanted to maximise this distance; and (2) as American movies and tv-shows are hugely popular in Sweden, it is conceivable that Swedes are familiar with several cultural differences between Sweden and the US, contributing to potential differences between the two imagined companies. That being said, it is evident that the interaction between language and culture is a complex one (as shown in the Ramírez-Esparza et al. (Citation2008) study reported in the Introduction). Thus, more systematically studying various combinations of both language and culture could add to the empirical base, yielding a deeper understanding of the effects of language and culture on personality dimensions in bilinguals. Finally, we want to point out the potential limitation of using questionnaire data, as these are potentially susceptible to cultural biases. However, the Big-5 personality inventory is robust and has been translated and validated across a large number of languages and cultures (e.g. Benet-Martinez & John Citation1998; Carciofo et al. Citation2016; Courtois et al. Citation2018). Nevertheless, future studies may wish to experimentally disentangle the effects of language and culture on personality by other means than personality questionnaires. For example, Rodríguez-Arauz et al. (Citation2017) investigated personality changes in Mexican-American bilinguals using open-ended personality self-descriptions in English and Spanish, which were then anaylsed qualitatively. Combining a similar approach whilst still keeping language and culture separate in a two by two fashion could shed light on the generalisability of these findings using other methods that may not share the same methodological limitations.

Another methodological aspect concerns the use of a between-subjects design, where participants were only presented with either the Swedish or the American company scenario, and completed the experiment either in Swedish or in English. This design was chosen for the current study to avoid potential carry-over effects from one of the conditions to another, but if carry-over effects could be avoided (e.g. by retesting the same participants in a different condition at a later date), more powerful designs (e.g. within-subjects design) could be implemented in future studies.

All in all, while several empirical questions remain to be investigated, this study is the first to attempt to investigate the separate contributions of language and culture in the literature of cultural frame switch and the foreign language effect in bilinguals.

Conclusions

The current study has, across two experiments, investigated the effect of language and culture as separate variables on bilinguals’ responses to the Big-5 Personality Inventory. The results highlight the potential need to separately measure and investigate the effect of language and culture in studies where either, or both, are manipulated but where it is difficult to disentangle the two. The general pattern of results, however, are in line with previous research on personality questionnaires in bilingual populations. As such, these results also provide a replication of previous research in a new population of bilinguals. On a more general level, our findings also show how volatile results on personality tests might be and how they can be affected by seemingly small manipulations where aspects such as different cultures or languages are made more or less salient, at least in bilingual populations.

Overall, this study contributes to the increasing literature on the how language, culture and emotion interact with one another, in an increasingly multilingual world.

Research data

Research data from this study is available upon request.

Acknowledgements

We thank Zulfikar Salman and Roosa Liukkonen for assisting with participant recruitment and data collection.

Disclosure statement

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