The TIE-93: a Facial Emotion Recognition Test Adapted for Culturally, Linguistically, and Educationally Diverse Alzheimer’s Dementia Patients in France

Abstract

Objective

Emotion recognition tests are essential for differential diagnostics when assessing patients with Alzheimer’s disease (AD) dementia. However, there remains a lack of emotion recognition tests appropriate for culturally and educationally diverse populations. The aim of this study was to develop an emotion recognition test (the TIE-93) appropriate for these populations. We then examined whether the TIE-93 could reduce emotion recognition performance differences between populations with a native French versus a culturally and educationally diverse background (participants who had immigrated to France). This was assessed by comparing performance between controls of each cultural group. We also assessed the effect of demographic variables on TIE-93 test performance and whether performance in an AD patient group was consistent with the research literature.

Methods

Fifty-seven patients with AD dementia and 240 healthy controls, from native French and culturally and educationally diverse backgrounds, were included in the study. The TIE-93 is composed of eight panels with photos of actors displaying six basic emotions. Participants were asked to identify which of the six facial expressions displayed matched an oral description of a context.

Results

When comparing French and culturally and educationally diverse controls, Quade’s ANCOVA revealed that there remained an effect of culture and education on TIE-93 test performance. Nonetheless, while controlling for years of education, age, sex, and cultural group, patients with AD dementia scored significantly more poorly than controls, specifically for most negative emotions.

Conclusion

The TIE-93 represents a first step toward developing appropriate emotion recognition tests for culturally and educationally diverse populations.

INTRODUCTION

Emotion recognition tests have recently been identified as essential for early detection and differential diagnostics within Alzheimer’s disease (AD) dementia (Torres Mendonça De Melo Fádel et al., 2019; García-Casal et al., 2019; Strijkert et al., 2022). As a fundamental component of social cognition, emotion recognition plays a critical role in enabling individuals to identify, interpret, and adapt to social situations, making its assessment a valuable predictor of social functioning (Ferreira et al., 2021; Trevisan & Birmingham, 2016). In AD, although deficits in episodic memory and orientation difficulties are most prominent, impairments in social behavior, emotion perception, and theory of mind may be more disturbing due to their associations with reduced quality of life and caregiver burden (Kessels et al., 2021). This is also reflected neuropathologically with degeneration initially affecting the hippocampi and the entorhinal and posterior cingulate cortices and then progressing to affect the entire temporal, parietal, and frontal cortices (Christidi et al., 2018; Frisoni et al., 2010). Identified as one of the six neurocognitive domains impaired in AD by the American Psychiatric Association, emotion recognition tests may help differentiate patients with AD dementia from those with mild cognitive impairment, major depressive disorder, and behavioral variant frontotemporal dementia (bvFTD; American Psychological Association, 2021; Bertoux et al., 2015; Fernandez-Rios et al., 2021; Bora et al., 2016; Strijkert et al., 2022; Ferrer-Cairols et al., 2023; Gressie et al., 2023; Park et al., 2017). However, testing emotion recognition, particularly within culturally, linguistically, and educationally diverse populations, remains challenging due to a scarcity of appropriately adapted tools (Matsumoto & Wilson, 2022; Franzen et al., 2021; see Bourdage et al., 2023 for a recent systematic review). Social cognition was highlighted as one of the most difficult cognitive domains to assess with current instruments in culturally diverse settings in Europe, stressing the urgent need for developments in this area (Franzen et al., 2021).

The effect of culture and education on neuropsychological tests, such as those of emotion recognition, is well documented, with education possibly being one of the most important cultural variables to consider when conducting neuropsychological assessments (Fernández & Evans, 2022; Fittipaldi et al., 2024). In general, neuropsychological tests have been developed for populations of the Global North, meaning countries that historically have had access to resources, less inequalities in living standards, and higher life expectancy due to economic and socio-political stability (Alladi & Hachinski, 2018). As a consequence, neuropsychological evaluations assume test-takers to have familiarity with a certain testing format, content, and structure (i.e., test-wiseness) and with the types of items and procedures used in the test (e.g., having items in a test familiar only to particular cultures, such as the “pretzel” item in the Boston Naming Test; Li et al., 2022; Fujii, 2018; Cipolotti & Warrington, 1995; Veliu & Leathem, 2017; Fernández & Evans, 2022). As a result, those who are culturally, linguistically, and educationally diverse (e.g., differing in literacy levels and educational content) tend to perform more poorly on tests, not necessarily due to their cognitive ability but due to these biases, possibly affecting diagnostic outcomes and healthcare management (Daugherty et al., 2017; Maillet et al., 2016; Nielsen et al., 2018; Olabarrieta-Landa et al., 2019; Pham et al., 2021). More specifically, the effect of education on neuropsychological assessments is more significant in those with lower education and progressively less impactful in moderate- to higher-educated populations due to a low ceiling of most neuropsychological assessments (Ardila, 1996; Nielsen & Waldemar, 2016; Tan et al., 2021; Franzen et al., 2023).

Furthermore, culture shapes the socialization of emotions in formal and non-formal educational systems, establishing cultural norms and leading to differences in emotion recognition tendencies among adults (Röttger-Rössler et al., 2015; Kitzmann, 2012). A recent investigation across 12 countries demonstrated a general effect of country of origin on emotion recognition test performance where 20.76% of variance was explained by this variable (Quesque et al., 2022, p. 676). This “considerable” variance is particularly significant given that it is 10 times higher than the effect of culture found on memory, attention, and spatial navigation, highlighting the important role culture plays specifically on measures of social cognition (Quesque et al., 2022, p. 676). Today, there exists a large body of research outlining cultural norms in emotional literacy, expression and perception, where differences can be found in: the words used to describe emotions, which emotions are appropriate to express socially, how directly emotions can be expressed, which information to take into consideration when attempting to identify an emotion, and how they are expressed facially (Mesquita & Frijda, 1992; Matsumoto, 2001; Mesquita & Walker, 2003; Van Hemert et al., 2007; De Leersnyder et al., 2015; Lim, 2016; Thong et al., 2023). Of note, researchers have also identified sex differences in emotion recognition ability, with women typically outperforming men, likely related to cultural norms where historically women have been expected to be caretakers and therefore primed to pay closer attention to the feelings of others (Abbruzzese et al., 2019; Hutchison & Gerstein, 2016; Wingenbach et al., 2018). In addition, older individuals tend to perform overall worse than younger individuals on tests of emotion recognition seemingly regardless of cultural heritage (Grainger et al., 2023; Hayes et al., 2020).

The vast effect of culture and education on emotion recognition and the limited number of tests appropriate for culturally, linguistically, and educationally diverse populations pose a serious challenge to neuropsychologists facing an increasing number of elderly culturally diverse patients with various literacy and education levels in clinics (Franzen et al., 2020; Franzen et al., 2021; Nielsen, 2022). Indeed, a recent systematic review outlined that there continue to be few tests of emotion recognition adapted for these populations, particularly tests that were also clinically validated (Bourdage et al., 2023). In addition to there being few emotion recognition tests, most of the tests currently available have been evaluated within predominantly young participants who are highly educated and cognitively healthy (Bourdage et al., 2023). There is therefore a need to develop emotion recognition tests appropriate for culturally, linguistically, and educationally diverse populations. Previous studies on culturally and educationally diverse groups have suggested that using tests with (a) ecological relevance, (b) an oral format (which limits the need for literacy skills and allows the test to be more easily assessed in different languages with an interpreter), and (c) pictorial stimuli (as opposed to stimuli containing letters and numbers, e.g., the Trail Making Test) may be more suitable for these populations (Franzen et al., 2020; Guo, 2022; Nielsen et al., 2019; Wong et al., 2024). Therefore, this study aimed to develop an emotion recognition test, named the Test d’Identification des Émotions Faciales (TIE-93), adapted for culturally, linguistically, and educationally diverse populations, by incorporating the previously mentioned suggestions, to reduce test performance differences between native French versus culturally diverse participants in healthy controls and patients with AD dementia. To assess whether the TIE-93 can reduce emotion recognition test performance differences between native French and culturally, linguistically, and educationally diverse groups, we aimed to (a) compare performances between both control groups (native French vs. culturally, linguistically, and educationally diverse group) while controlling for education, age, sex, and cultural group on the TIE-93 total score; (b) compare all controls to the patients with AD dementia group while controlling for education, age, sex, and cultural group on TIE-93 total score as well as each emotion sub-score; and (c) examine the effect of demographic variables such as cultural group, age, years of education, and sex on TIE-93 total score, controlling for patient/control status. We hypothesized that the native French and the culturally, linguistically, and educationally diverse control group would perform similarly on the TIE-93. We hypothesized that participants with AD dementia would have outcomes that reflect the AD research literature, that is, patients performing significantly more poorly than controls on the emotions of disgust, sadness, fear, and anger specifically. Finally, we also hypothesized that although there may be effects of age, sex, years of education, and cultural group on emotion recognition test performance, the size of the effect would be relatively small.

METHODS

TIE-93 Stimuli

The TIE-93 was inspired by Ekman’s emotion recognition test (Ekman et al., 1987). The stimuli are composed of photos of actors from the F.A.C.E.S. database (images from this database have been included in this study with author permission), with actors ranging in age with an even number of male versus female actors (Ebner et al., 2010; see Fig. 1). The database and thus the stimuli in the TIE-93 are all Caucasian actors. The photos of the actors, which are numbered and displayed simultaneously, facially depict: joy, disgust, anger, fear, neutrality, and sadness. Excluding the example panel, there are eight panels in total, each displaying the emotions in a random order to control for response biases (McGrath et al., 2010; Tang et al., 2014). Finally, the stimuli are in color, which assists emotion recognition ability (Ikeda, 2020).

TIE-93 Items

For the items of the TIE-93, a context from everyday life is orally given with each emotion cue. For each panel, the test-taker is asked to identify (by pointing or naming the number below the image) which image depicts the emotion the test administrator names within a context, for example, “show me in which image he is sad because he has just learned that his mother is very sick.” The six contexts were also chosen to be as relatable cross-culturally as possible; in addition to the one listed in the previous example, they are as follows: “He has just bitten into a rotten apple, he feels disgust,” “He is calm, he feels nothing special,” “He is scared, he is really afraid,” “He is happy, he has just won the lottery,” and, finally, “He is angry, he is upset.” The wording of the context varies slightly to match the age of the actor displaying the emotions, for example, if the actor is older the sadness emotional cue is for a friend rather than their mother. In addition, some wording is slightly altered to maintain test-taker engagement, for example, instead of “He is angry, he is upset” to say “She is fed up, she’s angry.”

Test Construction Considerations

Although Ekman’s six emotions (fear, anger, joy, sadness, disgust, and surprise) are suggested to be more or less universal, surprise was not included in the TIE-93 as research shows conflicting results regarding whether its detection is affected in patients with AD dementia (Ekman et al., 1987; Alfaisal & Aljanada, 2018; Elfenbein & Ambady, 2002; Pagani & Lombardi, 2000; Goodkind et al., 2015). Instead, a neutral facial expression was added as a control, as is typical for emotion recognition studies (Bänziger et al., 2009; Carrera-Levillain & Fernandez-Dols, 1994). Furthermore, older faces were specifically included given previous findings of a possible own-age bias in emotion recognition, where individuals have better recognition with similarly aged stimuli (Koen et al., 2021; Pizzio et al., 2022). A decision was also made to display the stimuli simultaneously to permit differentiation between emotions. This assists in emotion identification when similar valence emotions are displayed such as sadness and anger (Wang et al., 2020). Regarding the items, a context was given to each emotion cue as it helps mutual understanding of the definition of an emotion and assists in emotion identification, overall decreasing possible confusion around the abstract idea of an emotion (Abramson et al., 2021; Chen & Whitney, 2022; Msika et al., 2022). Indeed, a recent article highlighted the critique that Ekman’s emotion recognition test relies on abstract concepts of emotion and therefore leaving room for misinterpretation (Coppini et al., 2023).

Test Procedure and Administration

The test is administered orally, the test-taker is not required to read or write. The test administrator begins the test by explaining the procedure of the test to the test-taker and starting with an example, that is, that they will be shown facial images and that they are asked to match the emotionally cued situation with the face that expresses this emotion. During the example, the test administrator has the opportunity to correct the test-taker and to answer any questions they might have regarding the test procedure to ensure their understanding. Once the test administrator has confirmed the test-taker’s understanding of the procedure, they may begin testing, at which point they are no longer allowed to confirm or deny the accuracy of the response of the test-taker. Test administration time is approximately 8–10 min for controls and 10–12 min for patients. Correct answers sum to a total of eight for each emotion and a total overall score of forty-eight.

Participants

Regarding the recruitment of participants, which began in 2014, 33 patients with AD dementia and 93 healthy controls of French origin were recruited, along with 24 patients with AD dementia and 147 healthy controls from culturally diverse origins (meaning individuals who had immigrated to France and also had educational diversity) were included in the study. For the culturally, linguistically, and educationally diverse healthy controls, most were from the continent of Africa (n = 96, 65%), predominantly Algeria (n = 41, 27%), Tunisia (n = 17, 12%), and Morocco (n = 16, 11%; for a complete list of countries of origin, see Table 1). In total, 29 different countries were represented by the culturally, linguistically, and educationally diverse control group in this study. Culturally, linguistically, and educationally diverse controls were required to be able to speak, understand, read, and write in French as they needed to verbally discuss and read as well as sign, and confirm their understanding of the informed consent forms. All controls were recruited by students via the Université Paris Cité and by the Center of Health Exams (Centre d’Examen de Santé) in Bobigny as part of a voluntary free health checkup by the French National Health Service. For controls recruited by university students, students visited local community centers and utilized snowball sampling for recruitment. Controls then underwent a number of experimental tests including the TIE-93 if they reported a history absent of neurological or psychiatric challenges. For those recruited via the Center of Health Exams, controls were examined by a neurologist using the Diagnostic and Statistical Manual of Mental Disorders-V criteria to assess presence of neurocognitive or psychiatric diagnoses (and received subsequent medical referral if a diagnosis was suspected).

For culturally, linguistically, and educationally diverse patients with AD dementia, due to ethical legislation limitations in France, they were not asked for their country of origin but rather about their native language. The majority of patients reported Arabic (n = 10, 42%) or Creole (n = 5, 21%) as their native language (for a complete list of native languages, see Table 1). All patients were recruited by the memory clinic team at the Avicenne hospital in Bobigny, France, (which has received patients from culturally, linguistically, and educationally diverse backgrounds for nearly a century) where they underwent a neuropsychological evaluation with a clinical neuropsychologist, including a global cognitive function test (the Mini-Mental State Examination, MMSE) and other relevant testing (e.g., TNI-93) with the addition of the TIE-93 test (Dion et al., 2015; Kalafat et al., 2003; Maillet et al., 2016). All patients received a clinical diagnosis of AD, according to McKhann and colleagues’ (2011) diagnostic criteria, after a multidisciplinary team reviewed all data collected that included neuroimaging. In accordance with McKhann and colleagues’ (2011) guidelines, it was not recommended to collect biomarker data if the patient met core clinical criteria; consequently, only three AD patients from our study were asked to provide biomarkers. Of the patients diagnosed with AD dementia included in the study, four had the behavioral variant. Patients with concomitant cerebrovascular disease were excluded from this study (n = 2) as well as two patients with MCI possibly due to AD. The diagnosis of MCI versus AD was based on information on impairment in activities of daily living as reported in clinical interviews as there currently exist no validated tools for dementia staging in culturally, linguistically, and educationally diverse populations. Culturally, linguistically, and educationally diverse patients were asked to self-report whether they were able to speak and understand as well as read and write in French by responding either “yes,” “no,” or “somewhat.” Of the 24 patients, all except one reported that they were able to speak and understand French. One patient reported that they could somewhat speak and understand French. For reading and writing, 13 patients said “yes,” six said “no,” and six said “somewhat.” For all participants, education was measured by the age a participant ended their education. A participant was excluded from the study if they self-reported visual impairment or difficulties with comprehension or if these difficulties became apparent during the assessment. All participants performed the TIE-93 in French; this was a requirement to enter the study as the use of interpreters was not possible. All participants gave their informed and written consent to participate in the study, which followed the ethical guidelines of the Declaration of Helsinki and was approved by the Research Ethics Committee of the Université Paris Cité for control participants (registration number: 00012022-52) and the Comité Local d’Éthique pour la Recherche Clinique des HUPSSD Avicenne-Jean Verdier-René Muret for patient participants (CLEA-2023-n°306).

Statistical Analyses

The following statistics were performed using IBM SPSS Statistics version 28, with a significance level of .05. All scores are reported in medians and interquartile ranges. Due to minor deviations in the data from a normal distribution, Mann–Whitney U analyses were used to compare patients and controls within each cultural group (native French vs. culturally, linguistically, and educationally diverse) for age and years of education. For the first hypothesis, Quade’s ANCOVA (non-parametric) analysis was used to compare native French healthy controls to culturally, linguistically, and educationally diverse controls while controlling for the independent variables of years of education, age, and sex. For the second hypothesis, Quade’s ANCOVA was also used to compare patients to healthy controls, controlling for the variables of cultural group, years of education, age, and sex on the TIE-93 total score as well as for each emotion sub-score. Finally for our last hypothesis, for each predictor of the Quade’s ANCOVA, partial eta-squared (η²) effect sizes were calculated. For the purposes of ensuring accurate results given the lower statistical power of Quade’s ANCOVA, a generalized additive model was used to assess the relationship between the independent variables of cultural group, age, years of education, and sex on the TIE-93 total score while controlling for patient/control status. A generalized additive model was also used to assess the relationship between the independent variables of sex, MMSE score, age, and years of education on TIE-93 total score in French patients. Thin-plate regression splines were used to model non-linear relationships, with a basis dimension (k) set to 5 (to balance model flexibility and avoid overfitting) for smooth terms, and restricted maximum likelihood was applied for smoothness estimation. This analysis was performed in RStudio version 2024.09.0 with package “mgcv.” All p-values were adjusted using the Bonferroni correction method where significance levels were divided by the number of tests performed. The TIE-93’s internal consistency was assessed using the Spearman–Brown split-half reliability analysis and the Kuder–Richardson coefficient, appropriate for binary data.

RESULTS

All participants were able to complete the TIE-93 test; none reported having to stop due to the TIE-93’s level of difficulty or comprehension difficulties. Participant characteristics are summarized in Table 2. For an overview of test performance on the TIE-93 total score for native French and culturally, linguistically, and educationally diverse healthy controls and patients with AD dementia, please see Fig. 2.

Comparing Healthy Controls (Native French vs. Culturally, Linguistically, and Educationally Diverse) and Patients to Controls

Quade’s ANCOVA controlling for years of education, age, and sex demonstrated that native French healthy controls scored significantly higher on the TIE-93 total score (40 ± 5.9) compared to the culturally and educationally diverse controls (35 ± 7.9), F(1, 295) = 23.23, p < .001. Quade’s ANCOVA controlling for cultural group, years of education, age, and sex also revealed that patients scored significantly more poorly (34 ± 7.6) than controls (38.5 ± 7.3) on TIE-93 total score, F(1, 295) = 6.03, p = .02. Partial eta-squared (η²) effect sizes were calculated for each predictor and demonstrated large effects for cultural group (η² = 0.13, 95% CI [0.08, 1.00]), years of education (η² = 0.10, 95% CI [0.05, 1.00]), and age (η² = 0.13, 95% CI [0.08, 1.00]) and a small effect of sex (η² = 0.04, 95% CI [0.01, 1.00]). Quade’s ANCOVA controlling for the same independent variables was then conducted for each emotion sub-score. Patients scored significantly more poorly than controls specifically for the emotions of disgust [F(1, 295) = 10.15, p < .01], fear [F(1, 295) = 6.22, p = .01], and anger [F(1, 295) = 4.85, p = .03]. There were no significant differences found for happiness, sadness, or for the neutral emotion (see Fig. 3).

Generalized Additive Model

The results of the generalized additive model revealed similar results to the Quade’s ANCOVA calculations performed earlier: (a) the native French group scored significantly better than the culturally, linguistically, and educationally diverse group on TIE-93 total score, B = 5.16, t(295) = 6.21, p < .001; (b) controls scored significantly better than patients on TIE-93 total score, B = 3.32, t(295) = 3.36 t, p < .001; and (c) women scored significantly better than men on TIE-93 total score, B = 2.51, t(295) = 3.35, p < .001. The model explained approximately 32.9% of the variance in TIE-93 total scores, with cultural group accounting for 27% of the total variance. Another generalized additive model was run only with French patients while including MMSE scores to control for disease severity. However, the effect of MMSE scores on TIE-93 total score was non-significant. Interaction effects were analyzed for gender/age, gender/education, and age/education; however, none were significant. For additional information on the generalized additive model, including smooth terms, smooth term plots, and a table summarizing the results, please see the Supplementary material.

TIE-93 Reliability Analysis

The Spearman–Brown split-half reliability analysis for the TIE-93 was 0.85. For the Kuder–Richardson coefficient, item-level responses were not recorded for all controls; therefore, an analysis was conducted only on those with the necessary item-level detail, n = 54. In addition, when conducting analyses, the joy item for the fourth set of facial stimuli (an older male actor) was excluded as there was no response variance. Therefore, the Kuder–Richardson coefficient for the 47 items of the TIE-93 was 0.87.

DISCUSSION

This study aimed first to develop an emotion recognition test adapted for culturally, linguistically, and educationally diverse populations, named the TIE-93, by: (a) increasing ecological relevance by including a context to the emotion cue, (b) using an oral format thereby limiting the need for literacy skills, and (c) providing pictorial stimuli where facial images are displayed simultaneously to aid in emotion identification and limit misunderstanding. We then assessed whether the TIE-93 can reduce emotion recognition test performance differences between populations with a native French versus a culturally, linguistically, and educationally diverse background by comparing performance in healthy controls. We also assessed whether performance in an AD patient group was consistent with the research literature. Finally, we assessed the effect of cultural group, years of education, age, and sex while controlling for patient/control status on TIE-93 total score, hypothesizing that the effect of cultural group and years of education on TIE-93 would be relatively small. Overall, the TIE-93 shows good internal consistency, with one item removed from the analysis as it might have been too easy given that all participants responded correctly. Despite controlling for years of education, age, and sex, French controls still scored significantly higher on the TIE-93 than the culturally, linguistically, and educationally diverse controls, with the cultural group explaining 27% of total variance, similar to results obtained by Quesque and coworkers (2022) mentioned earlier. When comparing all patients to controls, our results were consistent with research literature. More specifically, our results demonstrated that patients scored significantly more poorly than controls on almost all negative emotions such as disgust, fear, and anger. As previously stated, these findings have been found in Global North populations and in an emotion recognition study with a Korean sample of patients with AD (Bertoux et al., 2015; Fernandez-Rios et al., 2021; Bora et al., 2016; Strijkert et al., 2022; Ferrer-Cairols et al., 2023; Gressie et al., 2023; Park et al., 2017). Contrary to expectations, there was no significant difference found for sadness, which may be due to limitations in test development. When assessing the effect of all independent variables (years of education, age, sex, and cultural group), all were significant. Although these effects were expected, the effect size for the variables of years of education and cultural group on TIE-93 test performance was not, demonstrating the difficulty of developing cross-cultural emotion recognition tests despite specific adaptations based on previous research.

This study has limitations that may explain the effect of years of education and cultural group on the TIE-93. First, test development did not strictly follow test development guidelines such as those from the International Test Commission (Muniz et al., 2013). For example, the items on the test were not reviewed by an expert panel, nor was the TIE-93 pre-tested to determine validity and reliability of items. This is a common limitation for test development in culturally and educationally diverse groups and often due to a lack of funding and/or due to time constraints, as was the case for this study; therefore, this study utilized a clinical-based approach (Bourdage et al., 2023). Ideally, to assess the validity of the TIE-93, performance comparisons would be made with other emotion recognition tests; however, there currently exist no appropriate tests of emotion recognition for culturally and educationally diverse groups, particularly for populations from the continent of Africa (Bourdage et al., 2023). Given the lack of pilot-testing limitation, the context given in the TIE-93 items may not be entirely recognizable cross-culturally; future research may be required. Another important limitation to note is that all facial stimuli are of Caucasian actors. Although this will present an intra-group disadvantage for culturally diverse test-takers, as outlined by Elfenbein and Ambady (2002), it was not possible to include facial stimuli from, for example, African cultures that represent a large majority of the migration population in France, due to the absence of validated facial stimuli databases with actors from diverse cultural backgrounds (Bourdage et al., 2023; Flahaux & De Haas, 2016). Additionally, the intra-group bias is reduced when cultures have greater exposure and familiarity with each other by living in the same nation, having physical proximity and frequent verbal communications (Elfenbein & Ambady, 2002). Given that the TIE-93 was to be assessed within the migration population in France, it was determined that, without a better option of including the appropriate culturally diverse stimuli, having the stimuli represent only Caucasian actors was an alternative solution. Furthermore, the use of the MMSE in culturally and educationally diverse patients is not ideal given the effects of culture and language on test performance; therefore, it was difficult to determine the accuracy of the completed items (Ng et al., 2018). It was for this reason that a generalized additive model with MMSE scores to control for disease severity was only possible in French patients, which was nevertheless non-significant. Furthermore, MMSE scores were not collected in healthy controls, limiting our ability to compare scores between controls and patients. Future studies would benefit from including the Rowland Universal Dementia Assessment Scale (RUDAS) as a global cognitive assessment as it is better adapted for educationally and culturally diverse populations and assessing healthy controls in addition to patients (Nielsen, 2022; Nielsen & Jørgensen, 2020). Also, the item-level detail for responses was not recorded for all controls, limiting the power of the internal consistency analysis. In addition, for the purposes of having sufficient power for analyses and small sample sizes, participants from largely differing cultures were grouped together limiting the ability to make comparisons between cultures.

This study also has strengths. For example, we have shown in a previous review that many researchers do not report their methodological approaches to test development for culturally, linguistically, and educationally diverse groups (Bourdage et al., 2023). This is a significant limitation as it does not allow for a quality assessment of works conducted to address the important resource gap for diverse populations. In response, this study outlines its test development procedure with both its limitations and strengths. In addition, as there is very little research conducted in test development for culturally, linguistically, and educationally diverse populations, this study’s primary strength is its contribution to a field of research still in its infancy. It is also a strength that the TIE-93 is a modification of Ekman’s emotion recognition test, a previously validated and globally used emotion recognition test, as it builds on its extensive work on “universal” emotions. Moreover, the TIE-93 also addresses some of Ekman’s test’s limitations, particularly regarding ecological relevance. This was done by modifying the test based on previous research outlining the possible best methods to create tests for culturally, linguistically, and educationally diverse populations and perhaps reducing the need to be test-wise (i.e., familiar and comfortable with cognitive testing). This resulted in a test that is entirely oral and has stimuli that are displayed simultaneously. In addition, the test addresses the abstract nature of Ekman’s emotion recognition test by providing a context to each emotion to reduce the likelihood of misinterpretation (Coppini et al., 2023). Also, the TIE-93 may be one of the few emotion recognition tests that include facial stimuli with elderly faces. Most emotion recognition assessments are tested on participants who are young, female, highly educated, and cognitively healthy. Therefore, it is a strength of this study that it included older participants of both sexes with varying levels of education and specifically included a group with cognitive impairment. Furthermore, this study contributes to the AD dementia literature by demonstrating similar patterns of response in emotion recognition in a culturally, linguistically, and educationally diverse group as previously reported in other cultural groups.

The TIE-93, in sum, represents a pertinent attempt at addressing the resource gap in emotion recognition tests for culturally, linguistically, and educationally diverse populations. In addition, this study contributes to the research literature by reporting methodological approaches, testing performance in culturally, linguistically, and educationally diverse populations of older age with a cognitive impairment group. This study also contributes to the AD literature by providing results from a culturally, linguistically, and educationally diverse migrant group in France. However, more research is required to: (a) further assess the limitations of emotion recognition testing in a culturally, linguistically, and educationally diverse context and determine when to use culturally individualized items in addressing these limitations and (b) understand the impact of education on the TIE-93 and how to better control for this important variable. Future research may also aim to further adapt the TIE-93 stimuli and items by including multicultural actors and making the items more applicable cross-culturally. Future research would also benefit from assessing whether the photos of the actors depicting the emotions are just as recognizable between emotions and whether there is an effect of younger versus older faces (as suggested by Chuang et al., 2021). In addition, it would be beneficial for future research to include the use of interpreters to examine the effect of language on TIE-93 test performance. It would also be pertinent for future research to assess global cognitive function in controls, which may be valuable to ensure the healthy functioning of a control and to assess variability of scores in this group when compared to the patient group. For generalizability, future research would benefit from larger patient sample sizes and to assess the TIE-93 in other international migration groups as the migration population in France may not be representative of other migration groups outside of France. Additionally, another recommendation would be to measure culture more consistently and include an assessment of acculturation such as the Brief Acculturation Scale for Hispanics (Norris et al., 1996). French and diverse groups were not educationally matched in this study, which would be recommended for future research and perhaps future research would also benefit from assessing education, as suggested by Fernàndez and Evans (2022), by a participant’s level of literacy rather than years of education. Future studies might also consider including specifically illiterate individuals as research has shown differences in neuropsychological test performance between literate and illiterate populations (Nielsen & Jørgensen, 2013). Finally, including patients at earlier stages of AD such as those with MCI due to AD and other etiologies such as frontotemporal dementia, Parkinson’s disease, and neurodevelopment or psychiatric conditions is also recommended. It would also be valuable to explore the performance of the TIE-93 in patients with localized lesions. Essentially, the TIE-93 holds potential as an emotion recognition tool and, with further research, may be clinically validated to assist in addressing the important resource gap for emotion recognition assessments for culturally, linguistically, and educationally diverse populations.

FUNDING

This work was supported by the IDEX Global Fellowship for author R.B. from Université Paris Cité. S.F. and J.P. are supported by ZonMW (#73305095007, #1051003210004) and Health Holland, Topsector Life Sciences & Health (PPP-allowance, #LSHM20106). S.F. and J.P. receive royalties on two neuropsychological tests (mVAT and FDT, Hogrefe). S.F. served as a consultant to Biogen in 2022.

CONFLICT OF INTEREST

None declared.

ACKNOWLEDGEMENTS

The authors would like to thank Hervé Le Clesiau who allowed the collection of data within the Centre d’Examen de Santé (Bobigny-93).

REFERENCES