Emotional fit in ratings versus language: A study of the emotional experiences of Dutch-speaking Belgians and Turkish migrants in Belgium

doi:10.21203/rs.3.rs-3744632/v1

Download PDF

Research Article

Emotional fit in ratings versus language: A study of the emotional experiences of Dutch-speaking Belgians and Turkish migrants in Belgium

https://doi.org/10.21203/rs.3.rs-3744632/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

When two or more people have the same emotion in relation to a situation, they are making meaning of it in the same way. Previous research has shown that the extent to which immigrant minorities share similar emotions with the receiving majority culture – termed ‘emotional fit’ – may bring better integration outcomes. Emotional fit is typically measured by comparing immigrants’ self-reported emotion intensity ratings to those of the average majority culture, at the expense of other aspects of emotion that might be relevant for meaning-making. To address these shortcomings, we implemented an alternative measure of emotional fit using natural language descriptions of emotional experience. Dutch-speaking Belgians and Turkish migrants in Belgium (Ns = 100) were interviewed about four recent emotional situations, two negative and two positive, for each one rating the intensity of their experience on a set of standardized emotion terms. Patterns of language used in the situation descriptions were more effective than patterns of intensity ratings in distinguishing between cultures, such that people fit better with their own culture than with the other, regardless of whether this was assessed within- or between-persons. Language- and rating-based fit measures were not positively correlated; rather, Turkish migrants’ rating fit with Belgian culture was negatively related to their language fit in negative situations, suggesting that even when they felt the same emotions, they were attending to different aspects of the experience. Future research is needed to further disentangle culture-specific patterns of emotional meaning-making and their implications for relevant outcomes for immigrant minorities.

emotion

culture

language

emotional fit

cultural fit

immigrant minority

Emotion is a form of meaning-making: a comparison between aspects of one’s current circumstances on the one hand, and a set of values and expectations on the other (Frijda, 1986; Mesquita, 2010). For example, being publicly reprimanded might engender anger when focusing on one’s rightfulness and valuing achievement of individual goals; or, it might engender embarrassment or respect when focusing on other people’s evaluation of the self or valuing protection of social rules. When two or more people have the same emotion in relation to a situation, they are making meaning of it in the same way – engaging in shared meaning-making. Accordingly, emotional similarity with others carries benefits for interpersonal interactions and relational health (Anderson et al., 2003; Gonzaga et al., 2007; Townsend et al., 2014). The same is true for emotional similarity with one’s surrounding cultural context, or emotional fit (De Leersnyder et al., 2014, 2015). Cultural influences on emotion lead to divergent patterns of emotional meaning-making for people with different cultural backgrounds (Mesquita, 2022; Mesquita et al., 2016), and the significance of these differences is perhaps starkest in the context of immigration, where immigrants must navigate emotional landscapes that are unfamiliar. Studying immigrants’ emotional fit with members of the local culture and how shared-meaning making is established between them can therefore grant insight into the barriers they face against their integration and how to mitigate those.

Emotional fit is typically assessed as the similarity between individuals’ emotional patterns and the average cultural pattern (e.g., De Leersnyder et al., 2011). To calculate this similarity, researchers rely on self-report measures in which participants rate the intensity of their experiences on a list of emotion words (e.g., “proud”, “embarrassed”, “grateful”). This approach offers summaries of emotional experience that are convenient, yet in two important respects narrow. First, it limits participants to a standard set of emotions on which to evaluate their experience. These pre-specified lists may prime certain types of emotional meaning-making by the words they include, and may miss important sources of individual and cultural variation by the words they do not (Li et al., 2020). Second, it focuses on the intensity of subjective feelings at the expense of other aspects that might be the focus of meaning-making. Knowing whether someone felt angry is not necessarily more relevant for interpersonal and intercultural relationships than knowing whether someone was alone versus in public, or whether they were attending to their own role versus that of others.

Natural language descriptions of emotional experiences offer a promising avenue for addressing both these shortcomings. Language serves as a window onto individuals' subjective experiences and is commonly used by researchers to access psychological states and processes (Tausczik & Pennebaker, 2010). There is ample evidence demonstrating that when people use similar patterns of language during interactions, they tend to experience more harmonious and fruitful exchanges (Ireland et al., 2011; Gonzales et al., 2010). It follows that fit in language use between an individual and the surrounding cultural context may be similarly impactful. Indeed, a recent study by Srivastava et al. (2018) demonstrated the utility of a language-based approach to cultural fit. Using e-mail communications from a mid-size technology firm, the research team derived patterns of language use with Linguistic Inquiry and Word Count (LIWC; Pennebaker et al., 2007), a software program that counts the percentage of words within a given document that correspond to a set of pre-defined categories ranging from basic linguistic features (e.g., pronouns, verb tenses, negations) to content categories (e.g., social, affective, cognitive), and found that employees’ (dis)similarity with the company norm (i.e., local organizational culture) was associated with outcomes such as retention and exit.

In the current study, we extended Srivastava et al.'s (2018) approach to create a measure of cultural fit in the domain of emotions using natural language descriptions of emotional experience (what we refer as language fit), and then compared this new measure with the traditional measure derived from intensity ratings (henceforth, rating fit). The data were collected with Dutch-speaking (majority culture) Belgians and first-generation Turkish migrants in Belgium as a part of a larger project investigating the role of emotions in the social integration and well-being of immigrant minorities. Turkish people constitute a substantial portion of the immigrant population in Belgium and there are well-documented differences in interpersonal perceptions, behaviors, and experiences between Turkish and Western European (e.g., Belgian) cultures (Mesquita, 2001; Rodriguez Mosquera et al., 2008; Uskul et al., 2012), providing a socially as well as psychologically valuable basis for comparing emotional meaning-making across these two groups.

In assessing an alternative measure of emotional fit across cultures, we first tested whether rating fit and language fit were actually cultural phenomena (Aim 1). We expected to find within- and between-culture differences in both fit measures, such that: (H1a) Turkish migrants’ fit with Turkish culture will be higher than their fit with Belgian culture; (H1b) Belgians’ fit with Belgian culture will be higher than their fit with Turkish culture; (H1c) Turkish migrants will fit with Turkish culture more than their Belgian counterparts; (H1d) Belgians will fit with Belgian culture more than their Turkish counterparts. Next, to compare our new measure of fit with the existing measure, we tested the association between language fit and rating fit within cultures and in the context of Turkish migrants’ fit with Belgian culture (Aim 2). We predicted that: (H2a) Belgians’ measures of fit with Belgian culture will be positively correlated; (H2b) Turkish migrants’ measures of fit with Turkish culture will be positively correlated; (H2c) Turkish migrants’ measures of fit with Belgian culture will be positively correlated.

The data analyzed in the current study were collected in the first phase of a large-scale, longitudinal study investigating the role of emotion in the social integration and well-being of immigrant minorities. As detailed below, participants were members of the Belgian (Dutch-speaking) majority in Flanders and recent Turkish migrants to Belgium. The study protocol was reviewed and approved by the KU Leuven Social and Society Ethics Committee (G-2020-2139-R3). Below we report aspects of the study relevant for the present analyses. All questionnaires administered during data collection, including those that were not used in the current study, are available on https://osf.io/jb74g/.

Participants

Participants were 100 Dutch-speaking Belgians (53 female, aged 18 to 71, M_age = 38,61, SD = 14.80) and 101 Turkish migrants (48 female, aged 19 to 52, M_age = 29,87, SD = 6.59) who completed both the initial survey and follow-up interview described below. Turkish migrants were randomly selected from a larger set of 280 Turkish participants for sample size comparability across cultural groups¹. A sensitivity analysis conducted in G*Power (Faul et al., 2007) showed that a sample size of 100 provided us with .80 power for ɑ < .05, two-tailed testing of a bivariate correlation to detect an effect size of r ≥ .27.

To be eligible, participants needed to be over 18 years old. Sample-specific requirements were as follows: for the Turkish sample, to be directly migrated from Turkey to Belgium between 2 months and 5 years prior to data collection; for the Belgian sample, to have no migration history dating back two generations and to have Dutch as their native language (or in case of French-Dutch bilinguals, to have been speaking Dutch since childhood). Turkish migrants arrived in Belgium an average of 19.90 months prior to data collection (range 3–53, SD = 12.91). Both samples were highly educated: 80 Belgian and 89 Turkish participants held a university or a higher educational degree, 19 Belgian and 10 Turkish participants were high school graduates. Turkish participants reported various reasons for migration, the most frequent of which were living in a safe place (n = 54), living a new life experience (n = 40), studying (n = 32), and being a political refugee (n = 32).

Procedure and Materials

Participants were recruited from communities in Flanders via personal networks and other convenience sampling strategies (e.g., distributing flyers, contacting organizations). All participants provided informed consent and took the study including all the materials in their native language. All materials were translated from English in the respective language (Dutch and Turkish) of the participant. Dutch materials were translated by the research team who were native Dutch-speakers. Turkish materials were first translated by a translation company; these translations were back-translated by two independent research assistants who were native speakers, with discrepancies resolved in group discussions. Data were collected in two phases: an initial survey, followed by a semi-structured interview. Participants received up to 50€ for completing both parts of the study (20€ for the interview for both groups; 15€ and 30€ for the survey for Belgians and Turkish sample respectively due to logistic differences explained below).

Initial Survey (Emotional Patterns Questionnaire)

After confirming they met the eligibility criteria, Belgian participants received a link to the online survey to complete on their own. Based on the experience with newcomers in previous studies (e.g., De Leersnyder et al., 2011; Mesquita, 2001), eligible Turkish newcomers were invited to complete the survey in the presence of a research assistant from their cultural background who read the questions out loud and recorded the responses in their native language. As part of this initial survey, all participants completed the Emotional Patterns Questionnaire (EPQ; De Leersnyder et al., 2011). The EPQ consists of prompts asking participants to recall and briefly describe four emotional situations (i.e., events, moments) from the past six months. The elicited situations differ in experienced pleasantness (i.e., valence; positive vs. negative) and interpersonal motive (i.e., motive; relationship-promoting vs. autonomy-promoting), two dimensions that have been found to structure emotional experience across cultures (De Leersnyder et al., 2015; Kitayama et al., 2006). The four situational prompts can be found in Table S1 in Supplementary Materials.

All participants were asked to report all four situations; however, one Turkish and one Belgian participant failed to report a negative autonomy-promoting situation, and an additional Belgian participant did not report the two relationship-promoting situations. Therefore, we collected 200 positive relationship-promoting (101 Turkish, 99 Belgian), 200 negative relationship-promoting (101 Turkish, 99 Belgian), 201 positive autonomy-promoting (101 Turkish, 100 Belgian), and 199 negative autonomy-promoting (100 Turkish, 99 Belgian) situation descriptions.

After describing each situation, participants were asked to rate the intensity of their emotions at the time of the event on 20 different emotion terms using a 5-point Likert scale (1 = Not at all; 5 = Very strongly). The emotion terms covered a wide range of emotions representing the four emotion types, or combinations of valence and motive (positive relationship-promoting: respectful towards the others, close to the others, helpful towards the others; positive autonomy-promoting: proud, happy, elated [excited]; negative relationship-promoting: guilty, ashamed, indebted; negative autonomy-promoting: frustrated, angry, resentful), complemented by emotion terms that are frequently used in literature (good, bad, calm, worried, nervous, fearful, sad, surprised).

Semi-structured Interviews

After completing the survey, participants were asked if they were interested in taking part in an additional interview. Those who agreed were asked to describe the four EPQ situations in greater detail. The interviews were guided by a set of questions about specific aspects of the situation, such as the time and the place, other people involved, subjective feelings, bodily sensations, and what they and others did. These aspects are commonly cited as core features or components of emotion (e.g., Frijda, 1986), and were intended to help participants to relive and actively make meaning of their experience. The interview scheme can be found in the Appendix of the Supplementary Materials.

Interview duration ranged between 30 and 60 minutes. Although the original plan was face-to-face data collection, due to COVID-19 regulations the majority of interviews were held online and recorded via Microsoft Teams or Zoom. All interviews were conducted and transcribed by research assistants fluent in the respective languages.

Data Preparation

Prompt Check

We checked whether participants recalled situations that corresponded with the prompts in two ways. We started by checking, in a two-step process, whether the reported situations met the valence of the prompt. In the first step, two independent coders (native speakers of the respective languages) read the open-ended descriptions of the emotional situations and made a three-level assessment on valence (positive, negative, unclear). Both disagreements and the descriptions evaluated as unclear were discussed among the two coders and a final decision was made as to whether the situation was positive, negative, or unclear. As a second step, we checked whether the ratings of the situation as good and bad corresponded with the valence prompts. We considered a situation as matching when at least one criterion (either the coders’ assessment, or the rating as good or bad) was clearly met. Therefore, even if the situation met the valence of the prompt on one criterion, and was neutral (i.e., a description coded as unclear or equal ratings of good and bad) on the other, we considered the prompt met. However, when one criterion contradicted the expected valence, we excluded the data from the analyses concerning that situation type. When both of these criteria were neutral, we checked the ratings for happy and angry, under the expectation that happy (angry) ratings would be higher for positive (negative) situations. Based on these valence-based checks, data were excluded for 9 situations² in the Turkish sample, and 12 situations³ in the Belgian sample.

Next, we confirmed that participants recalled situations that corresponded with the prompts in terms of both interpersonal motive and valence, using the intensity ratings. For each culture, we calculated the average ratings on the four emotion types (positive relationship-promoting, positive autonomy-promoting, negative relationship-promoting, negative autonomy-promoting) and compared them across situations using repeated measures ANOVAs. Post-hoc comparisons with Bonferroni corrections revealed that in all situations, and for both Belgian and Turkish participants, the ratings for the emotion type matching the prompt were higher than the ratings for the other emotion types (see Table S2). This means that participants followed the prompts well, as confirmed by manual examination of the content and by an analysis of the corresponding intensity ratings.

Cultural Equivalence

We checked for measurement equivalence in the emotion terms across the two cultural groups using Simultaneous Component Analysis (SCA; De Roover et al., 2012). This method allowed us to examine the structural similarities of multiple variables (i.e., emotion terms) across multiple participant blocks (i.e., cultural groups). We conducted SCA on the emotion ratings (18 terms, excluding good and bad⁴) from all four situations, with two blocks (Turkish, Belgian), and allowing for five components (one up to four expected components; i.e., emotion types based on valence and interpersonal motive). A three-component solution was the best fit to the data for both groups, explaining 64.17% of the total variance. Based on this solution, we decided to exclude three emotion terms from further analyses. Two of these terms (frustrated and calm) loaded on different components across cultures. The third term (surprised) loaded on a component that included negatively valenced, autonomy-promoting emotions (e.g., anger). This was unexpected given that surprised has previously loaded together with positive autonomy-promoting emotions (e.g., proud; e.g., De Leersnyder et al., 2011, Study 2), or has loaded differently across cultural groups (e.g., De Leersnyder et al., 2011, Study 1). With comparability in mind, we decided to exclude surprised from the present calculations, leaving us with 15 emotion terms (respectful towards the others, close to the others, helpful towards the others, proud, happy, elated [excited], guilty, ashamed, indebted, angry, resentful, worried, nervous, fearful, sad).

Cultural fit in emotion intensity ratings (rating fit)

We used the emotion intensity ratings from the EPQ to calculate the extent to which individuals’ patterns of emotional experience fit with the average patterns of the respective culture in a situation type. As detailed further below, we analyzed EPQ and interview data separately for each of the four situation types (positive relationship-promoting, positive autonomy-promoting, negative relationship-promoting, and negative autonomy-promoting).

Rating fit was calculated following De Leersnyder et al. (2011). That is, we created profiles of the 15 emotion intensity ratings for each participant, separately for each of the four situations. Then, we averaged all the participants’ profiles in a culture to obtain the average emotional pattern of that culture in the respective situation type. Lastly, we correlated each participant’s profile with the average cultural profile⁵. The resulting Fisher-transformed correlation coefficients were used as estimates of rating fit, representing the similarity of an individual’s emotional pattern to the average emotional pattern in that culture within the given situation type. We calculated four types of rating fit in this way: Belgian majorities’ fit to the Belgian pattern, Turkish migrants’ fit to the Turkish pattern, Turkish migrants’ fit to the Belgian pattern, and Belgian majorities’ fit to the Turkish pattern.

Cultural fit in language of emotion descriptions (language fit)

We used the transcripts from the semi-structured interviews to calculate the extent to which the patterns of language people used to describe the elicited situations fit with those used by the respective culture. We began by manually splitting the transcripts by situation type (four documents per participant), and then used freely-available software (ConverSplitterPlus, 2022) to generate documents with only participant (and no interviewer) speech. Since we were interested in cross-cultural comparisons of language (Turkish migrants’ fit with Belgian culture), we translated all documents in English in order to avoid language structure and (Dutch and Turkish) LIWC dictionary-specific properties from biasing the results. Transcripts were automatically translated using DeepL (March, 2022) for the Dutch and DeepL (September, 2023) for the Turkish transcripts.

Patterns of language use were quantified using LIWC. We used the 2007 version of LIWC (Pennebaker et al., 2007), which includes over 60 categories organized in a structure of parent (e.g., Social processes) and sub-categories (e.g., Family, Friends, Humans). We submitted the translated situation-level, participant-only documents to the English LIWC2007. To ensure that Dutch and Turkish interviews are covered by the LIWC to similar extent, we compared the dictionary coverage across the two samples An independent samples t-test conducted on situation-level data showed no difference, t(801) = .02, p = .99, so that the English LIWC captured 95.26% of the words used in both of the translated Dutch and Turkish transcripts.

To calculate cultural fit in language, we followed the strategy we used with the emotion intensity ratings, this time using LIWC category scores to create profiles. Following Srivastava et al. (2018), we included all LIWC categories except for punctuations, which provided us with 64 word categories (see Table S3 in Supplementary Materials for the complete list). Using these categories, we created a profile for each participant for each situation and then averaged these profiles within each culture to obtain the cultural pattern of language use in the respective situation. We correlated each participant’s profile with the average cultural profile and Fisher-transformed the correlation coefficients to obtain language fit, representing the similarity of an individual’s language pattern to the average language pattern in that culture. As with emotional fit, language fit was computed at the situation level.

Before testing our hypotheses, we checked if we could create composite estimates of rating fit and language fit by averaging the situation-level coefficients for each measure. Reliability analyses of fit estimates across four situation types showed poor reliability for rating fit, with Cronbach’s α of .25 for rating fit with the Turkish culture (.20 in the Turkish migrant sample; .22 in the Belgian sample) and of .17 for rating fit with the Belgian culture (.15 in the Turkish migrant sample; .22 in the Belgian sample). Cronbach’s alpha for language fit with the Turkish culture was .79 (.61 in the Turkish migrant sample; .70 in the Belgian sample), and for language fit with Belgian culture was .86 (.68 in both cultural samples). Although reliabilities for language fit were acceptable, we conducted all analyses at the situation level to keep our analyses and interpretations comparable across the two fit measures.

Aim 1: Is Emotional Fit a Cultural Phenomenon?

To meet our first aim, that is to answer whether measures of emotional fit reflect cultural phenomena, we conducted within- and between-culture comparisons of rating fit and language fit across the four situation types.

Rating fit

We tested whether people had higher rating fit with their own cultural group than with the other cultural group using paired-sample t-tests. Results offered partial support for H1a and H1b for rating fit. Turkish migrants indeed fit better with the Turkish culture in three situation types (positive relationship-promoting, negative relationship-promoting, and negative autonomy-promoting; ps ≤ .002), but not in the positive autonomy-promoting situations (t(99) = 1.80, p = .08). A similar pattern was observed for Belgians, where they fit better with the Belgian culture in the same three situations, ps ≤ .02, but not in positive autonomy-promoting situations (t(96) = .04, p = .97).

We also checked whether people have higher rating fit with their own cultural group than people from the other cultural group using independent samples t-tests. Partially supporting H1c and H1d for rating fit, results revealed expected patterns only in negative situations and for some cultural comparisons. Specifically, Belgians fit better with Belgian culture (M = .81, SD = .35) than Turkish migrants did (M = .66, SD = .26) in negative relationship-promoting situations, t(188) = .2.64, p = .009; and Turkish migrants fit better with the Turkish culture (M = .94, SD = .41) than Belgians did (M = .83, SD = .38) in negative autonomy-promoting situations, t(191) = .3.66, p < .001. Results did not provide evidence for cultural differences in positive situations, ps > .13. All comparisons can be seen in Table 1.

Table 1
Within- and Between-Culture Comparisons in Measures of Emotional fit
Rating fit
Within culture comparisons
Situation type	Cultural group	Mean (SD)				t (df)
Situation type	Cultural group	Fit to Turkish		Fit to Belgian		t (df)
Positive Relationship-promoting	Turkish	1.35 (.54)		1.33 (.53)		3.11** (100)
Positive Relationship-promoting	Belgian	1.32 (.52)		1.35 (.55)		2.41* (96)
Positive Autonomy-promoting	Turkish	1.41 (.53)		1.40 (.52)		1.80 (99)
Positive Autonomy-promoting	Belgian	1.29 (.55)		1.29 (.55)		.04 (96)
Negative Relationship-promoting	Turkish	.75 (.38)		.65 (.46)		4.02*** (95)
Negative Relationship-promoting	Belgian	.66 (.26)		.81 (.35)		6.24*** (93)
Negative Autonomy-promoting	Turkish	.94 (.41)		.83 (.38)		5.72*** (96)
Negative Autonomy-promoting	Belgian	.74 (.31)		.84 (.35)		5.57*** (95)
Between culture comparisons
Situation type	Fit type		Turkish - Belgian (SE)		t (df)
Positive Relationship-promoting		Fit to Turkish	.03 (.08)			.41 (196)
Positive Relationship-promoting		Fit to Belgian	− .03 (.08)			− .33 (196)
Positive Autonomy-promoting		Fit to Turkish	.12 (.08)			1.54 (195)
Positive Autonomy-promoting		Fit to Belgian	.11 (.08)			1.40 (195)
Negative Relationship-promoting		Fit to Turkish	.09 (.05)			1.81 (188)
Negative Relationship-promoting		Fit to Belgian	− .16 (.06)			-2.64** (188)
Negative Autonomy-promoting		Fit to Turkish	.19 (.05)			3.66*** (191)
Negative Autonomy-promoting		Fit to Belgian	− .01 (.05)			− .18 (191)
Language fit
Within culture comparisons
Situation type	Cultural group	Mean (SD)				t (df)
Situation type	Cultural group	Fit to Turkish		Fit to Belgian		t (df)
Positive Relationship-promoting	Turkish	2.78 (.24)		2.46 (.18)		18.50*** (99)
Positive Relationship-promoting	Belgian	2.47 (.18)		2.79 (.24)		18.99*** (96)
Positive Autonomy-promoting	Turkish	2.68 (.25)		2.43 (.18)		15.76*** (99)
Positive Autonomy-promoting	Belgian	2.50 (.18)		2.77 (.22)		17.29*** (96)
Negative Relationship-promoting	Turkish	2.81 (.25)		2.50 (.19)		17.48*** (95)
Negative Relationship-promoting	Belgian	2.53 (.21)		2.86 (.26)		18.40*** (94)
Negative Autonomy-promoting	Turkish	2.84 (.27)		2.52 (.20)		15.47*** (96)
Negative Autonomy-promoting	Belgian	2.55 (.22)		2.89 (.26)		17.36*** (95)
Between culture comparisons
Situation type	Fit type		Turkish - Belgian (SE)		t (df)
Positive Relationship-promoting	Fit to Turkish		.31 (.03)		10.17*** (195)
Positive Relationship-promoting	Fit to Belgian		− .33 (.03)		-10.94*** (195)
Positive Autonomy-promoting	Fit to Turkish		.18 (.03)		5.82*** (105)
Positive Autonomy-promoting	Fit to Belgian		− .34 (.03)		-12.07*** (195)
Negative Relationship-promoting	Fit to Turkish		.28 (.03)		8.33*** (189)
Negative Relationship-promoting	Fit to Belgian		− .36 (.03)		-10.90*** (189)
Negative Autonomy-promoting	Fit to Turkish		.30 (.04)		8.39*** (191)
Negative Autonomy-promoting	Fit to Belgian		− .37 (.03)		-11.08*** (191)
Note. All analyses were conducted with 2-tailed testing. p < .05, p < .01, **p < .001.

Language fit

Paired-sample t-tests for within-culture comparisons revealed that in all four situation types, Turkish migrants’ language fit better with the Turkish culture, ps < .001, and Belgians’ language fit better with the Belgian culture, ps < .001. Thus, H1a and H1b were fully supported for language fit.

Independent samples t-tests for between-culture comparisons revealed that in all four situation types, Turkish migrants fit better with the Turkish culture than Belgians did, ps < .001, and Belgians fit better with the Belgian culture than Turkish migrants did, ps < .001. H1c and H1d were also fully supported for language fit.

All comparisons can be seen in Table 1.

Exploratory analyses

For both cultural groups and for each measure of emotional fit, we examined the associations between participants’ fit with their own culture and with the other to see whether this would provide insight into why H1a-d were met for language fit but only partially met for rating fit. The reasoning is that the higher the association between fit with Belgian culture and fit with Turkish culture, the less likely it is that the fit measure captures cultural differences.

For rating fit, Pearson’s correlations showed that fit with one culture was strongly associated with fit with the other (rs ranging from .76 to .99; see Table S4), such that when people fit well with Turkish culture it is highly likely that they also fit well with Belgian culture (and vice versa). A similar pattern was observed for language fit, but to a lesser extent, such that fit with Turkish culture and fit with Belgian culture were positively associated (rs ranging from .65 to .79, see Table S4).

Lastly, we tested whether the strength of association between fit with one culture and fit with the other culture statistically differed between rating fit and language fit. For this purpose, we compared correlation coefficients for the two measures (correlation between rating fit with Turkish culture and rating fit with Belgian culture on one hand and the correlation between language fit with Turkish culture and language fit with Belgian culture on the other hand) within cultures across situations using confidence intervals (Diedenhofen & Musch, 2015; Zou, 2007). Results revealed that, overall, the association between fit with Turkish culture and fit with Belgian culture was higher for rating fit than for language fit (see Table S4).

Aim 2: Is Language Fit Associated with Rating Fit?

For our second aim, that is to test the association between language fit and rating fit, we first checked this relationship in Belgian and Turkish migrant samples separately. Then, to test the language fit approach in the context of immigration, we examined if Turkish migrants’ language fit with the Belgian culture converged with their rating fit with the Belgian culture. In all cases, we ran bivariate Pearson’s correlations between language fit and rating fit using a two-tailed test of significance at ɑ = .05.

Within-culture comparisons

Rating fit and language fit were not consistently related to each other in either cultural group. The association was strongest, albeit nonsignificant, in negative relationship-promoting situations; notably, it was also negative (r(94) = − .19, p = .07, 95% CI [-.377, .015] in the Belgian sample; r(96) = − .15, p = .14, 95% CI [-.343, .049] in the Turkish migrant sample). A correlation matrix of the relationship between measures of emotional fit across situations can be seen in Table 2.

Table 2

Correlations between Language Fit and Rating Fit
Situation Type	r	p	df
Within-culture: Belgians
Positive relationship-promoting	.10	.35	97
Positive autonomy-promoting	.08	.42	97
Negative relationship-promoting	− .19	.07	94
Negative autonomy-promoting	.14	.18	96
Within-culture: Turkish migrants
Positive relationship-promoting	− .09	.40	100
Positive autonomy-promoting	.04	.70	100
Negative relationship-promoting	− .15	.14	96
Negative autonomy-promoting	.02	.86	97
Between-culture: Turkish migrants’ fit with Belgians
Positive relationship-promoting	− .07	.50	100
Positive autonomy-promoting	− .05	.61	100
Negative relationship-promoting	− .26	.01	96
Negative autonomy-promoting	− .25	.02	97
Note. The associations that are statistically significant (p < .05) are represented in bold.

Between-culture comparisons

Turkish migrants’ language fit and rating fit with the Belgian culture were negatively associated in negative situations, such that when they reported feelings similar to the average Belgian feelings during unpleasant events, the language features they used to talk about these events differed from those of the average Belgian (r(96) = − .26, p = .01, 95% CI [-.437, − .063] for negative relationship-promoting events; r(97) = − .25, p = .02, 95% CI [-.425, − .049] for negative autonomy-promoting events). There was no relationship between Turkish migrant’s language fit and rating fit with the Belgian culture in positive situations (see Table 2).

We investigated cultural fit in emotions using natural language descriptions, as well as intensity ratings, of emotional experiences collected from Dutch-speaking Belgians and Turkish migrants in Belgium. We included language in study of (dis)similarities in emotions because of its potential to provide insight into emotional meaning-making beyond the intensity rating scales provide. We began by assessing whether measures of emotional fit (both rating fit and language fit) reflected cultural differences in how participants described four types of emotion-eliciting situations varying in valence (positive vs. negative) and interpersonal motive (relationship- vs. autonomy-promoting). Consistent with our expectations, we found that cultural groups could be clearly differentiated based on their patterns of language use – for example, Turkish migrants’ language use was more similar to other Turkish migrants than it was to Belgians. Surprisingly, these cultural differences were weaker and less consistent for the traditional rating fit. We next assessed the relationship between the two measures of fit. Contrary to our expectations, we did not find a positive association between language fit and rating fit. In fact, in some analyses the association was negative, particularly when we compared Turkish migrants’ ways of fitting with Belgians in negatively-valenced situations.

The present analyses suggest that language fit may be more reliable than rating fit in differentiating cultural groups. To follow up on the inconsistency of expected cultural differences in rating fit (H1a-d), we examined the associations between fitting in with one culture and with the other. For both rating fit and language fit, people who fit better with the Turkish culture also fit better with the Belgian culture. However, our results showed that positive associations between cultural fits were stronger for intensity ratings than for language use. This imbalance might be partly explained by methodological differences in calculating the two types of fit. While we include all types of language properties (i.e., all LIWC dictionaries) in estimating language fit, we remove the emotion items that cluster differently across cultures when estimating rating fit. This preprocessing step ensures cultural equivalence of the concepts we are comparing, but leaves out items that are cross-culturally different in meaning. It may thus lead to an underestimation of the actual differences that does not occur in calculating language fit. Also, rating fit might be more susceptible to prompt-specific constraints than language fit, because the emotion items used in the rating task more closely relate to the prompts (e.g. valence) than most language categories used to estimate language fit (exceptions are positive vs. negative affect categories).

The lack of a consistent relationship between the fit measures suggests that self-reported intensity ratings and natural language descriptions capture distinct facets of emotional meaning-making. Even if two people give a similar pattern of responses on rating scales, they may diverge in how they conceptualize the emotional experiences connected to those ratingswhich may reveal itself in language. The words people use can be seen as indicators of their mental attention (Boyd & Schwartz, 2021) – in descriptions of emotional experience, perhaps the features (bodily sensations, cognitive processes, involvement of others, etc.) that are being foregrounded. For instance, looking at the words Chinese Americans used in their free descriptions of emotional experiences, Tsai and colleagues (2004) found that migrants’ focus on social and somatic aspects varied as a function of their acculturation orientation. Accordingly, the present findings might be better explained by considering language fit to represent similarity in how attention is deployed across a wide range of emotional features. These features may not be equally foregrounded when people rate emotion words.

The negative association between Turkish migrants’ rating fit and language fit with Belgian culture in negative situations is noteworthy. This finding suggests that when Turkish migrants reported feelings similar to those of an average Belgian, they tended to focus on different features of the experience in their descriptions. For example, when Turkish migrants described offenses, they might have been feeling angry to a similar extent as Belgians but differentially attending to aspects of the situation, for instance paying more attention to social concerns, resulting in different patterns of language, for instance more social words (Tsai et al., 2004). The types of events described might also differ, for example, offenses might involve a close or a distant other, which lead to different appraisal patterns, e.g. attribution of blame, across cultures (Boiger et al., 2018). The appraisal of blame, and who to blame, might reveal itself in language, in form of differing (co)-occurrence of certain words, such as pronouns and intentional verbs (Windsor et al., 2014). One limitation of the current approach to measuring cultural fit in language is that the properties causing differences in fit are not directly observable; only overall estimates of fit are compared. Although including all types of words to estimate language fit allowed us to take a comprehensive, holistic perspective on patterns of language use, it does not allow us to examine which aspects of experience are driving emotional meaning-making. Future research is needed to explore these possibilities further; for instance, by manually coding relevant components of emotions (e.g., social concerns, or appraisals of self- and other-blame), by examining the types of events using automated approaches such as topic modelling (Blei, 2012), and by doing a targeted examination of language properties and their co-occurence to uncover the associations between the types of emotional situations, appraisals, features of the experience being attended to, and subjective feelings.

A multimethod approach to emotional fit may provide valuable insights into the intricate and diverse nature of emotional meaning-making in intercultural contexts, and language is a powerful tool by which we can examine multiple aspects of the emotional experience. As a first step, we demonstrated that cultural differences can be captured by the language used to talk about emotional experience; people have higher language fit with their own culture than the other, showing face validity of our approach. Our tests of convergent validity were not met – language fit and rating fit were not positively related – leaving questions for future research about why these measures deviate and what they each represent. At the same time, the language data revealed differences in migrants’ emotional meaning-making that were not captured by the intensity ratings alone. Future research will need to assess predictive validity by examining whether language fit is differentially associated with relational, psychological, and societal outcomes of interest in the context of immigration.

Acknowledgements: The authors are grateful to Tom Verthé and Alba Jasini for their major role in preparation and data collection for this study.

Funding: K.H. was supported by the Research Foundation – Flanders (12A3923N) and by a Marie Skłodowska-Curie Individual Fellowship from the European Commission (892379) under the European Union’s Horizon 2020 research and innovation program. R.S.Ş. and B.M. were supported by the European Research Council (Advanced Grant 834587), also under Horizon 2020. This paper reflects only the authors’ views; the European Commission and European Research Council are not liable for any use that may be made of the contained information.

Conflict of Interests: There is no conflict of interest with respect to the research, authorship, and/or publication of this article.

Ethics Approval: This study was performed in line with the principles of the Declaration of Helsinki. Approval was granted by the KU Leuven Social and Society Ethics Committee (G-2020-2139-R3).

Consent to participate: Informed consent was obtained from all individual participants included in the study.

Material, Data and Code Availability: The anonymized data used in the current study is a subset of the 1st wave of the Newcomers Study, which is a large-scale, longitudinal study investigating the role of emotion in the social integration and well-being of immigrant minorities. As this manuscript is the first to be submitted that uses this dataset, the whole set of materials are made available at https://osf.io/jb74g/ along with the portion of the data necessary to replicate the reported analyses and analytic code for fit score calculations. To protect participant privacy, an following our ethics guidance, the demographic variables and the raw language data are not publicly available.

Author Contribution: B.M. wrote the grant including the current study, designed the study, and provided the data. R.S.Ş and K.H. made the analysis plan, R.S.Ş analyzed the data. R.S.Ş and K.H. drafted the manuscript. All authors provided revisions and approved the final version of the manuscript.

Anderson, C., Keltner, D., & John, O. P. (2003). Emotional convergence between people over time. Journal of Personality and Social Psychology, 84(5), 1054-1068. https://doi.org/10.1037/0022-3514.84.5.1054
Babcock, M. J., Ta, V. P., & Ickes, W. (2014). Latent semantic similarity and language style matching in initial dyadic interactions. Journal of Language and Social Psychology, 33(1), 78–88. https://doi.org/10.1177/0261927X13499331
Blei, D. M. (2012). Probabilistic topic models. Communications of the ACM, 55(4), 77–84. https://doi.org/10.1145/2133806.2133826
Boiger, M., Ceulemans, E., De Leersnyder, J., Uchida, Y., Norasakkunkit, V., & Mesquita, B. (2018). Beyond essentialism: Cultural differences in emotions revisited. Emotion, 18(8), 1142-1162. https://doi.org/10.1037/emo0000390
De Leersnyder, J., Kim, H., & Mesquita, B. (2015). Feeling right is feeling good: Psychological well-being and emotional fit with culture in autonomy-versus relatedness-promoting situations. Frontiers in Psychology, 6, 630. https://doi.org/10.3389/fpsyg.2015.00630
De Leersnyder, J., Mesquita, B., Kim, H., Eom, K., & Choi, H. (2014). Emotional fit with culture: A predictor of individual differences in relational well-being. Emotion, 14(2), 241-245. https://doi.org/10.1037/a0035296
De Leersnyder, J., Mesquita, B., & Kim, H. S. (2011). Where do my emotions belong? A study of immigrants’ emotional acculturation. Personality and Social Psychology Bulletin, 37(4), 451–463. https://doi.org/10.1177/0146167211399103
De Roover, K., Ceulemans, E., & Timmerman, M. E. (2012). How to perform multiblock component analysis in practice. Behavior Research Methods, 44(1), 41–56. https://doi.org/10.3758/s13428-011-0129-1
DeepL Translate (n.d.). https://www.deepl.com/translator
Diedenhofen, B., & Musch, J. (2015). cocor: A comprehensive solution for the statistical comparison of correlations. PloS One, 10(4), e0121945. https://doi.org/10.1371/journal.pone.0121945
Faul, F., Erdfelder, E., Lang, A.-G., & Buchner, A. (2007). G* Power 3: A flexible statistical power analysis program for the social, behavioral, and biomedical sciences. Behavior Research Methods, 39(2),
175–191. https://doi.org/10.3758/bf03193146
Frijda, N. H. (1986). The Emotions. Cambridge University Press.
Gonzaga, G. C., Campos, B., & Bradbury, T. (2007). Similarity, convergence, and relationship satisfaction in dating and married couples. Journal of Personality and Social Psychology, 93(1), 34-48.
https://doi.org/ 10.1037/0022-3514.93.1.34
Gonzales, A. L., Hancock, J. T., & Pennebaker, J. W. (2010). Language style matching as a predictor of social dynamics in small groups. Communication Research, 37(1), 3–19. https://doi.org/10.1177/0093650209351468
Ireland, M. E., Slatcher, R. B., Eastwick, P. W., Scissors, L. E., Finkel, E. J., & Pennebaker, J. W. (2011). Language style matching predicts relationship initiation and stability. Psychological Science, 22(1),
39–44. https://doi.org/10.1177/0956797610392928
Kitayama, S., Mesquita, B., & Karasawa, M. (2006). Cultural affordances and emotional experience: Socially engaging and disengaging emotions in Japan and the United States. Journal of Personality and Social Psychology, 91(5), 890-903. https://doi.org/10.1037/0022-3514.91.5.890
Mesquita, B. (2001). Emotions in collectivist and individualist contexts. Journal of Personality and Social Psychology, 80(1), 68-74. https://doi.org/10.1037//0022-3514.80.1.68
Mesquita, B. (2010). Emoting: A Contextualized Process. In The mind in context (pp. 83–104). The Guilford Press.
Mesquita, B. (2022). Between us: How cultures create emotions. WW Norton & Company.
Mesquita, B., Boiger, M., & De Leersnyder, J. (2016). The cultural construction of emotions. Current Opinion in Psychology, 8, 31–36. https://doi.org/10.1016/j.copsyc.2015.09.015
Pennebaker, J., Chung, C., Ireland, M., Gonzales, A., & Booth, R. (2007). The Development and Psychometric Properties of LIWC2007.
Rodriguez Mosquera, P. M., Fischer, A. H., Manstead, A. S., & Zaalberg, R. (2008). Attack, disapproval, or withdrawal? The role of honour in anger and shame responses to being insulted. Cognition and Emotion, 22(8), 1471–1498. https://doi.org/10.1080/02699930701822272
Srivastava, S. B., Goldberg, A., Manian, V. G., & Potts, C. (2018). Enculturation trajectories: Language, cultural adaptation, and individual outcomes in organizations. Management Science, 64(3), 1348–1364. https://doi.org/10.1287/mnsc.2016.2671
Tausczik, Y. R., & Pennebaker, J. W. (2010). The psychological meaning of words: LIWC and computerized text analysis methods. Journal of Language and Social Psychology, 29(1), 24–54. https://doi.org/10.1177/0261927X09351676
Townsend, S. S., Kim, H. S., & Mesquita, B. (2014). Are you feeling what I’m feeling? Emotional similarity buffers stress. Social Psychological and Personality Science, 5(5), 526–533. https://doi.org/10.1177/1948550613511499
Tsai, J. L., Simeonova, D. I., & Watanabe, J. T. (2004). Somatic and social: Chinese Americans talk about emotion. Personality and Social Psychology Bulletin, 30(9), 1226–1238. https://doi.org/10.1177/0146167204264014
Uskul, A. K., Cross, S. E., Sunbay, Z., Gercek-Swing, B., & Ataca, B. (2012). Honor bound: The cultural construction of honor in Turkey and the Northern United States. Journal of Cross-Cultural Psychology, 43(7), 1131–1151. https://doi.org/10.1177/0022022111422258
Windsor, L. C., Dowell, N., & Graesser, A. (2014). The language of autocrats: Leaders’ language in natural disaster crises. Risk, Hazards & Crisis in Public Policy, 5(4), 446–467. https://doi.org/10.1002/rhc3.12068
Zou, G. Y. (2007). Toward using confidence intervals to compare correlations. Psychological Methods, 12(4), 399-413. https://doi.org/10.1037/1082-989X.12.4.399

A subset of 100 Turkish participants was selected using random number generation. Survey data were missing for one of these participants, so another participant with full data was randomly selected as a replacement. Afterwards, the survey data were located for the initial participant, bringing the final sample size to 101 participants.
5 negative relationship-promoting, 1 positive autonomy-promoting, and 3 negative autonomy-promoting situations.
2 positive relationship-promoting, 4 negative relationship-promoting, 3 positive autonomy-promoting, 3 negative autonomy-promoting situations.
Items good and bad were only included for valence check purposes.
When fit is calculated within culture (e.g., fit of Belgians to the average Belgian pattern), the individual’s own profile was excluded from the average cultural profile to avoid inflating concordance.

No competing interests reported.

SupplementaryMaterialsAffectiveScienceSencanetal.docx

Download PDF

Version 1

posted

You are reading this latest preprint version

Emotional fit in ratings versus language: A study of the emotional experiences of Dutch-speaking Belgians and Turkish migrants in Belgium

Status:

Version 1

Abstract

Introduction

Method

Participants

Procedure and Materials

Initial Survey (Emotional Patterns Questionnaire)

Semi-structured Interviews

Data Preparation

Prompt Check

Cultural Equivalence

Cultural fit in emotion intensity ratings (rating fit)

Results

Aim 1: Is Emotional Fit a Cultural Phenomenon?

Rating fit

Language fit

Exploratory analyses

Aim 2: Is Language Fit Associated with Rating Fit?

Within-culture comparisons

Between-culture comparisons

Discussion

Declarations

References

Footnotes

Additional Declarations

Supplementary Files

Status:

Version 1