Adaptation and Validation of a French Version of the Griffith Empathy Measure

Although empathy is considered to be necessary for prosocial behavior and moral development (e.g. Denham, 2007, 2017; Eisenberg, 2005), there is no single definition of it that is universally agreed, because empathy involves several processes (Decety & Cowell, 2014). In developmental psychology and psychopathology, theoretical models of empathy have evolved from unitary to multi-component conceptions, influencing how empathic skills are defined (Barbot et al., 2022; Nader-Grosbois & Simon, 2023) and how they are measured in children (for a review, see Simon & Nader-Grosbois, 2023b). In the developmental model designed by Hoffman (1987), empathy referred to the state of a subject that is induced by the state of another person, to that it corresponds more closely than to the state of the subject (Hoffman, 2000, p.7). Four stages of empathy were distinguished: (1) “Emotion contagion”, displayed in infants and representing an automatic affective reaction to another person’s apparent emotion or distress; (2) “Attention to others’ feeling”, displayed from the age of one, and reflecting an interest in the emotions or distress expressed by others; (3) “Prosocial actions”, which begin in the second year of life, when the child feels concerned about others’ emotions and becomes able to help, support or comfort other people who display distress; and (4) “Empathy for another’s life condition”, from late childhood, corresponding to empathic skills concerning the general context of another person’s life. In multidimensional models influenced by studies conducted in developmental psychopathology and neuropsychology, affective, cognitive and behavioral empathy were differentiated. The affective component refers to an emotional response and a capacity to share others’ emotions appropriately, while the cognitive component concerns the ability to understand others’ emotions or distress, by taking others’ perspective and decoding socioemotional cues in social situations; the behavioral component is displayed through prosocial actions (e.g., Barbot et al., 2022; Blair, 2005; Cuff et al., 2016; Davis, 1983; Decety, 2015; Dvash & Shamay-Tsoory, 2014; Eisenberg & Fabes, 1990; Eisenberg, 2005; Nader-Grosbois & Simon, 2023). In the literature, cognitive empathy is often quasi-synonymous with affective ToM (e.g., Blair, 2005; Nader-Grosbois & Simon, 2023). The distinction between affective and cognitive empathy was supported by behavioral and neuroimaging studies (e.g., Decety & Cowell, 2014; Decety, et al., 2018; Dvash & Shamay-Tsoory, 2014). In an analogous way, ToM also encompasses multiple processes that could be activated differently depending on social situations and required mental states – either emotions and desires (affective ToM) or knowledge, beliefs, intentions, etc. (cognitive ToM) – in others (Flavell, 1999; Nader-Grosbois, 2011). Intricate links between cognitive empathy and ToM have been reported (e.g., Eisenberg et al., 2006; Nader-Grosbois & Simon, 2023).

Depending on the theoretical background and the definition of empathy used, different measures have been designed to study how empathy develops from infancy to adolescence, to highlight protective versus risk factors that could impact its evolution and to investigate links with other domains (Eisenberg & Fabes, 1990; for a survey, see Simon & Nader-Grosbois, 2023b).

First, observational designs have been set up to decode facial affect or gestural reactions to others’ emotions and distress in situations where adults simulate the expression of these emotions, to identify the level of empathy in infants, toddlers or preschoolers (e.g. Kochanska et al., 2010; Sigman et al. 1992; Spinrad & Stifter, 2006; Skwerer & Tager‐Flusberg, 2016; Zahn-Waxler et al., 1992a, b). These coding methods are expensive to use with large samples and offer a limited approach to empathy skills depending on very specific target contexts.

Second, videos or pictures illustrating stories have been used to present hypothetical situations in order to elicit empathetic reactions or to invite the child to respond in tasks. These designs result either in a general score for empathy (e.g. Affective Situations Test for Empathy, Feshbach & Roe, 1968; Empathy Measure for Preschoolers, EMP, Sezov, 2002; Empathy Continuum, EC, Strayer, 1993) or in scores for affective, cognitive and behavioral empathy (e.g., Kids’ Empathic Development Scale, KEDS, Reid et al., 2013; Empathy Task, Bensalah et al., 2016a, b). Although these tasks are very relevant, there are potential biases due to the motivational factor or to the cognitive and verbal abilities that the children generally need to use.

Third, to remedy these methodological biases, an assessment in which adults report on children’s empathy through a questionnaire can be useful. In a unitary approach, some other-reported questionnaires give an overall score for empathy, for example, the Empathy Scale of Infant–Toddler Social and Emotional Assessment (ITSEA, Carter et al., 2003; French adaptation ESEF, Bracha et al., 2007), the Empathy Scale of My Child (Kochanska et al., 1994), the Empathy Quotient – Children (EQ-C, Auyeung et al., 2009) or the Adolescent Empathy Quotient (Auyeung et al., 2012). The Empathy and Theory of Mind Scale (EToMS, Wang & Wang, 2015) offers the possibility of obtaining a general score for empathy and a score for ToM skills used for good and bad purposes.

Fourth, based on multidimensional conceptions of empathy, questionnaires have been designed to provide differentiated scores. Based on the developmental model of Hoffman (1987, 2000), the Empathy Questionnaire (EmQue, Rieffe et al., 2010; French adaptation EmQue-vf, Simon et al., in revision) assesses adults’ perceptions of a child’s ability to empathize in daily life, with reference to the three first stages of empathy. It produces three scores, for emotion contagion, attention to others’ feelings and prosocial actions. In addition, questionnaires assess different dimensions, particularly with regard to affective and cognitive empathy. Self-report questionnaires are completed by children at school age: for example, the adaptation of the Interpersonal Reactivity Index (IRI) for children (Litvack-Miller et al., 1997), the Bryant’s Index of Empathy (Bryant, 1982), the Basic Empathy Scale (BES) for adolescents (Jolliffe & Farrington, 2006), a French adaptation of BES for children (Bensalah et al., 2016a, b), the Feeling & Thinking (F&T, Garton & Gringart, 2005), or the Cognitive, Affective and Somatic Empathy Scales (CASES, Raine & Chen, 2018). There is also an other-reported questionnaire, the Griffith Empathy Measure, (GEM, Dadds et al., 2008), which provides two scores for affective empathy and cognitive empathy and can be completed by adults (parents or teachers) with regard to children at preschool and school age.

In research and in intervention programs, it is important to use a reliable questionnaire that can be completed by multiple informants who are involved in the child’s life, especially when the child is young and/or has conditions that reduce his or her ability to understand items and to use self-report forms. As Dadds et al. (2008, p. 112) mention, the use of self-reports on empathy is problematic: children under the age of about 8 years have insufficient cognitive and/or verbal abilities to report on internal states, and there a lack of convergence in older children between their reports about affective empathy, their scores on picture-story indices and their prosocial behavior. This study therefore focused on the validation of a French adaptation of the GEM. The differentiation of empathy components is very important to examine at preschool and elementary school age, specific characteristics of typically developing (TD) children and of atypically developing children, notably those presenting externalized or internalized behavior disorders or autism spectrum disorders (ASD) or intellectual disabilities (Blair, 2005; Nader-Grosbois & Simon, 2023; Simon & Nader-Grosbois, 2023a).

Based on a multidimensional design, the Griffith Empathy Measure (GEM, Dadds et al., 2008). It adapts Bryant’s Index of Empathy (Bryant, 1982) which was a self-reported questionnaire for children and adolescents, turning it into an other-reported questionnaire for parents by reformulating items in third-person format. For each of the 23 items, the parent rates the degree of agreement with statements about behaviors that could be displayed by the child or adolescent in everyday social situations, on a 9-point Likert scale, ranging from “strongly disagree” to “strongly agree”. The total score can be computed by averaging the adult’s rating in all items. Two subscores for affective empathy and cognitive empathy can be calculated by averaging the rating for the respective items included in each factor. Higher scores reflect a greater degree of empathy.

The validation study for the GEM (Dadds et al., 2008) showed its good reliability and validity. The principal components analysis with Oblimin rotation with Kaiser normalization, applied to a sample of 2612 children from 4 to 16 years of age, identified a two-factor structure for cognitive and affective empathy, accounting for 22.32% and 15.03% of variance respectively. A confirmatory factor analysis was performed, with maximum likelihood estimation to check model fit and to test this two-factor structure across age groups and genders. This analysis showed that the GEM can be used for a wide range of ages and for both genders as an overall scale of empathy using the 23 items (α = 0.81); alternately, a cognitive empathy subscale (α = 0.62, 6 items) and an affective empathy subscale (α = 0.83, 9 items) can be differentiated after omitting items that load on both subscales. Main effects for three age groups (4–6, 7–10, 11–16 years) were obtained for the total and cognitive empathy scores, but not for the affective empathy score. Significant effects for gender were found, with girls showing higher overall, affective and cognitive empathy. Based on 127 parents’ GEM ratings, test–retest stability over 6 months was confirmed. Both mothers’ and fathers’ ratings of 155 children aged 5–12 years showed an acceptable agreement for the three scores. The convergent validity was confirmed by positive significant correlations (p < 0.01) between scores in 49 adolescents’ self-ratings of empathy in the Bryant Index for Empathy and mothers’ ratings for the three GEM scores. Behavioral observations of 28 boys from 6 to 12 years who participated in three activities with a pet mouse were rated by observers from videotapes with respect to three dimensions: (1) nurturing (caring, empathic, gentle behavior), (2) cruelty (careless and/or aggressive behavior likely to distress the animal) and (3) engagement (active verbal and/or nonverbal involvement with the animal). The results showed the expected convergence between cruel behavior toward the pet mouse and low GEM affective empathy, and conversely between nurturing behavior and high GEM affective empathy. Moreover, 23 children from 7 to 12 years, including 15 with oppositional defiant disorder and 8 with internalizing problems, were assessed with the GEM and the Interpersonal Response Task (IRT, Hawes & Dadds, 2004). Mothers’ ratings of empathy in the GEM (particularly the total and affective scores) were linked with both the forfeiting of rewards and decision reaction times in IRT. Good psychometric properties of GEM were thus confirmed.

The GEM allows multi-informant data to be collected about empathy from mothers, fathers, teachers or other professionals close to the child. It overcomes the limitations and potential biases of performance measures and can complement observations and results obtained by the latter. Given its interest, the GEM has been translated and adapted into several languages, including German, Chinese, and Spanish, but not yet into French. In the literature, the GEM has been used in numerous recent studies, showing its applicability to healthy, TD children or atypically developing children in diverse social environments.

Several studies have used the GEM with TD children at preschool age or at school age, particularly between 6 and 12 years (Bigelow et al., 2021; Dawel et al., 2015; Decety, et al., 2018; Demedardi et al., 2021; Gevaux et al., 2020; Guo & Feng, 2017; Kohls et al., 2009; Lavertu et al., 2022; McDonald et al., 2015; Rasmussen et al., 2016; Rong et al., 2022; Simon & Nader-Grosbois, 2021; Tuerk et al., 2021; Vera-Estay et al., 2016; Zhang & Wang, 2019). Links between their empathic skills, personality traits and responsiveness to social rewards have been explored (Kohls et al., 2009). Dawel et al. (2015) examined how affective and cognitive empathy in TD children correlated with their ability to discern authenticity in emotional facial expressions (happy, sad or scared). Some studies have analyzed links between these children’s empathy profiles and interactions with animals (Dadds et al., 2008; McDonald et al., 2015). Simon and Nader-Grosbois (2021) examined mothers’ and fathers’ perceptions of cognitive and affective empathy in preschoolers, and their variability according to age, gender and personality. Links between both dimensions of empathy and the children’s social adjustment were also investigated. Demedardi et al. (2021) studied whether emotional understanding and empathy in children were predictors of prosocial lying. Associations between video game playing and empathy, prosocial behavior, social adaptive skills or social behavior problems were explored by Lavertu et al. (2022). Vera-Estay et al. (2016) investigated the potential moderating and mediating role of executive functions and social cognition (empathy and ToM) in the link between children’s age and moral maturity. Tuerk et al. (2021) tested the biopsychosocial SOcio‐Cognitive Integration of Abilities modeL integrating temperament, executive functioning, communicative skills and social cognition (including empathy). Associations between children’s empathy and their reactions to unequal distributions of resources between two puppets and their parents’ teaching of “just world beliefs” were studied by Gevaux et al. (2020). Bigelow et al. (2021) examined whether children’s language mediate the relationship between age and both cognitive and affective ToM, and also affective or cognitive empathy. In a developmental neuroscience study, Decety et al. (2018) analyzed the electrophysiological responses when preschoolers perceived painful versus neutral stimuli and the association between these responses and perspective taking and empathic concern (assessed by parents who completed the GEM), as well as their relation to parental empathy and children’s own prosocial behavior. Guo and Feng (2017) investigated the links between parenting styles (emotional warmth, rejection) as perceived by children and their altruistic behavior and the intervening role of their empathy (rated by their mothers in the GEM). Efficiency of an intervention toward preschoolers was evaluated by means of the GEM as pre- and post-tests (Rasmussen et al., 2016).

In addition, numerous comparative studies of atypically developing children have explored aspects of affective and cognitive empathy using the GEM. These often involve comparisons between one or two clinical groups and TD children, mainly at school age and in a few studies at preschool age.

Most of the studies using the GEM have investigated affective and cognitive empathy in children or adolescents presenting externalized behavioral disorders, including complex conduct problems such as aggressive behavior (Dadds et al., 2012; Hawes et al., 2020; Deschamps et al., 2018; Fleming et al., 2022; Malcolm-Smith et al., 2015), disruptive behavior disorder (DBD) (Datyner et al., 2016; Deschamps et al., 2015), or opposition-defiant disorder (ODD) (Hawes et al., 2020; O’Kearney et al., 2017; Pasalich et al., 2014), or presenting attention deficit or hyperactivity disorders (ADHD) (Deschamps et al., 2015; Gumustas et al. 2017; Kohls et al., 2014). The main aim of these studies was to identify specific weaknesses and strengths of children’s empathy according to their differentiated clinical symptoms of externalized behavior disorders, in order to improve differential diagnostic processes and provide guidelines for clinical intervention. Some studies have also focused on links between such children’s affective or cognitive empathy and their difficulties in prosocial behavior (Deschamps et al., 2015) or their proactive and reactive aggression (Hawes et al., 2020; Deschamps et al., 2018; Malcolm-Smith et al., 2015). Other studies have examined links between difficulties in emotion recognition, understanding of emotions, emotion perspective taking or ToM, and affective or cognitive empathy (Lui et al., 2016; Malcolm-Smith et al., 2015; O’ Kearney et al., 2017). Kohls et al. (2014) compared empathy (measured with the GEM), affiliative tendency, interpersonal competences and neural activation in response to both social reward types in children with ADHD or ASD and TD children. The GEM was used at pre- and post-test in children with conduct problems to assess the efficiency of emotion recognition training (Dadds et al., 2012), of a Coaching and Rewarding Emotional Skills (CARES) program (Datyner et al., 2016) or of Parent–Child Interaction Therapy (PCIT) (Fleming et al., 2022).

Several studies have assessed affective and cognitive empathy, by means of the GEM, in children with “callous-unemotional traits” (CU) (Dadds et al., 2009, 2012; Datyner et al., 2016; Georgiou et al., 2019a, b; Fleming et al., 2022; Georgiou et al., 2019a, b; Hartman et al., 2019; Hawes & Dadds, 2012; Hawes et al., 2020; Kimonis et al., 2016; Lui et al., 2016; McDonald et al., 2018; O’Kearney et al., 2017; Pasalich et al., 2014). Links between the intensity of CU traits and affective or cognitive empathy in preschoolers and children at school age have been examined (Dadds et al., 2009; Hawes et al., 2020; Georgiou et al., 2019a, b; Hartman et al., 2019; Kimonis et al., 2016; Lui et al., 2016; Malcolm-Smith et al., 2015; McDonald et al., 2018; O’Kearney et al., 2017). For example, empathy and antisocial behavior in these children have been studied (Dadds et al., 2009; Kimonis et al., 2016). The relationships between animal abuse, CU traits and empathy have been explored in children whose mothers were exposed to partner violence (Hartman et al., 2019).

In addition, several studies of children with ASD have used the GEM to examine their affective and cognitive empathy and shown differentiated links with the severity of ASD symptoms (Andrews et al., 2013; Alkire et al., 2021; Deschamps et al., 2014; Georgiou et al., 2019a, b; Greimel et al., 2011; Jin et al., 2020; Kirst et al., 2022; Kohls et al., 2014; Shi et al., 2020; Soorya et al., 2015). Recently, there has been a growing consensus regarding this imbalance between cognitive and affective empathy in children and adolescents with ASD, based on the use of the GEM (e.g., Georgiou et al., 2019a, b). The role of mutual emotional experiences in social-interactive success, ToM skills and social processing in association with empathy in school-aged children with ASD has been explored (Alkire et al., 2021). Some studies have investigated links between empathy and executive functions in children with high-functioning ASD (Jin et al., 2020). A few studies have directly examined the neural mechanisms of impaired empathy in clinical groups, including ASD, using a measure of Regions-of-interest-based functional connectivity and the GEM (e.g. Shi et al., 2020). The GEM has been used at pre- and post-test, to test the efficiency of training programs in empathy and emotional skills in children with Asperger’s syndrome or ASD: for example, a cognitive–behavioral intervention program aimed at improving affectionate communication and friendship skills (Andrews et al., 2013), the Seaver-NETT program (Nonverbal communication, Emotion recognition, and Theory of mind Training) (Soorya et al., 2015) or the parent-assisted serious game Zirkus Empathico (Kirst et al., 2022).

Some studies have used the GEM with children with intellectual disabilities (Williams syndrome, Osorio et al., 2019; X-fragile syndrome, Miller et al., 2022; Down syndrome and nonspecific intellectual disabilities, Simon & Nader-Grosbois, submitted). Miller et al. (2022) tested whether empathy is impaired and associated with anxiety in girls with fragile X syndrome. Osorio et al. (2019) analyzed the associations between altered patterns of fetal testosterone and hypersociability, affective and cognitive empathy, anxiety and autistic symptoms in children with Williams’ syndrome. Simon and Nader-Grosbois (submitted) compared affective and cognitive empathy in children with intellectual disabilities and TD children matched for chronological age or developmental age, to examine developmental delay or deficit hypotheses of empathy. Links between empathy, ToM, emotion regulation and social adjustment have been studied in children with intellectual disabilities (Simon & Nader-Grosbois, in preparation-c). More rarely, the GEM has been used with children presenting internalized problems such as anxiety (e. g., O’Kearney et al., 2017), with mild traumatic brain injury (mTBI) (D’Hondt et al., 2017), or at risk of abuse and maternal maltreatment (Meidan & Uzefovsky, 2020).

516 children aged from 3 to 12 years (M = 6.42; SD = 2.65) and their parents took part in this study. Recruitment was carried out through the principals of kindergartens and elementary schools in the French-speaking area of Belgium. Table 1 shows information about socioeconomic status, including parents’ education level and family income, the children’s living arrangements and the presence of siblings. The parents indicated their level of education on a seven-point scale from low (elementary school not completed) to high (university). The mothers had an average level corresponding to Master’s degree, and fathers, to Bachelor degree. In terms of family income, the parents reported mean salaries and benefits on a 13-point subscale from 0–500 euros to more than 6000 euros. The mean income reported by parents corresponded to 3000 to 3500 euros per month. The great majority of children lived with their two parents. Almost half of the children had only one brother or sister.

Table 2 presents means and standard deviations for scores in GEM-vf, ToMI-1-vf, ToM task Battery-vf and ERC-vf, for all samples, by gender and age.

A principal component analysis with oblimin rotation was performed, using IBM SPSS Statistics 27.0, to determine the factorial structure of GEM-vf. In accordance with the components of affective and cognitive empathy found by Dadds et al. (2008), the number of factors was forced at 2. A first analysis with the 23 items was computed and showed an adequate adjustment of the data (KMO = 0.817; Bartlett’s Test of Sphericity: p < 0.001). The loading of item 3 was less than 0.3. Four items (7, 9, 16 and 22) loaded on the wrong factor and the content of item 21 did not seem to refer to affective or cognitive empathy. Consequently, these six items were removed from the analysis. The second exploratory factor analysis also showed a proper adjustment (KMO = 0.827; Bartlett’s Test of Sphericity: p < 0.001). The first factor, referring to affective empathy, explained 26.7% of the variance, while the second, cognitive empathy, explained 11% of the variance. The final model (see Table 3) contained 17 items, distributed as follows: 13 items for affective empathy and 4 for cognitive empathy.

Internal consistency was assessed to test the reliability of GEM-vf. The Cronbach’s alpha should be greater than or equal to 0.7 to be adequate and the average inter-item correlations should be in the range of 0.15 to 0.50, as recommended by Clark and Watson (1995). Table 4 shows Cronbach’s alphas and inter-item correlations for the overall items and items of both components. For the total and affective empathy, internal consistency were very good and met the two criterion. For cognitive empathy, Cronbach’s alpha was lower than expected and inter-item correlations met the criterion.

Pearson’s correlations were computed between age and the three scores in GEM-vf. Significant and positive correlations were found between age and total empathy (r = 0.262, p < 0.001), affective empathy (r = 0.237, p < 0.001) and cognitive empathy (r = 0.180, p < 0.001). A one-way MANOVA was run to check whether empathy differed depending on age group (see Table 2 for means and standard deviations, and Figs. 1, 2, 3). A main effect of group was demonstrated (Pillai’s F = 9.768, p < 0.001, η_p² = 0.037). The between-subjects test found significant differences in affective empathy (F = 16.738, p < 0.001, η_p² = 0.061) and in cognitive empathy (F = 6.131, p = 0.002, η_p² = 0.023). Bonferroni’s post-hoc highlighted a difference in affective empathy between the 3–5 years and 6–8 years groups (mean difference = -0.52, p < 0.001) as well as between the 3–5 years and 9–12 years groups (mean difference = -0.58, p < 0.001). For cognitive empathy, a difference was obtained between the 3–5 years and 9–12 years groups (mean difference = -0.55, p = 0.002). The one-way ANOVA that was run for the total empathy score showed a main effect of age group (F = 16.127, p < 0.001, η² = 0.059). Moreover, Bonferroni’s post-hoc showed differences in total empathy between the 3–5 years and 6–8 years groups (mean difference = -0.42, p < 0.001) and between the 3–5 years and 9–12 years groups (mean difference = -0.55, p < 0.001).

To examine gender difference in affective, cognitive and total empathy scores, a one-way MANOVA was conducted (see Table 2 for means and standard deviations); this showed no main effect of gender (Pillai’s F = 2.666, p = 0.07, η² = 0.010). However, between-subjects tests demonstrated a significant difference depending on gender in affective empathy (F = 4.559, p = 0.033, η² = 0.009), in the sense that girls had better affective empathy skills than boys; they did not differ significantly for cognitive empathy (F = 0.197, p = 0.658, η² = 0.000) or for total empathy (F = 0.708, p = 0.4, η² = 0.001).

Convergent validity was tested by Pearson’s coefficients of correlation analyses between scores in GEM-vf and in ToMI-1-vf, ToM-task Battery-vf and ERC-vf for all samples. Table 5 presents the results of correlations between all these measures. The total, affective and cognitive empathy scores in GEM-vf were positively and significantly linked with the three subscores (thoughts, beliefs, socioemotional), the total score in ToMI-1-vf (p < 0.001) and the subscores (affective, cognitive, mixed ToM) and the total score of ToM-task Battery-vf (p between 0.001 and 0.042). Moreover, the three scores in empathy in GEM-vf correlated positively and significantly with the score in emotion regulation in ERC-vf (p < 0.001). Significant and negative correlations were obtained between the three scores in GEM-vf and the score in emotion dysregulation in ERC-vf (p < 0.001).

Table 6 presents means and standard deviations for scores in the measures of empathy (GEM-vf and EmQue-vf) in the subsample of 299 preschoolers. Table 7 presents the results of Pearson’s correlations between scores in the two empathy measures. Positive and significant correlations were obtained between the scores for total empathy and affective empathy in GEM-vf and the scores for attention to others’ feelings and prosocial actions in EmQue-vf (p < 0.001). The score for cognitive empathy was significantly and positively related to the score for prosocial actions (p = 0.002), but the correlation with attention to others’ emotions was close to nul.