English (pdf)
Article in xml format
Article references
Curriculum ScienTI
Automatic translation
Permalink
Introduction
Empathy is a social ability that involves the interpersonal sharing of needs, experiences, and desires (Hall & Schwartz, 2019). Although it is considered an essential element regarding relationships, there have been registered differences concerning the number of dimensions that it is composed of. It is often described as being formed by two dimensions: affective and cognitive (Azevedo et al., 2018); whereas other authors mention a third dimension, named behavioral (Clark et al., 2018; Falcone et al., 2008).
Among the most employed definitions of empathy and its development in studies, the theoretical proposal formulated by Davis (1980) stands out by using other individuals’ observed experiences to define the cited social ability. Therefore, the cognitive domain, named perspective taking, refers to the ability to recognize others’ points of view, while the affective dimension, characterized as empathic concern, involves the feelings of consideration, worry, or compassion for others’ emotional states. Moreover, Davis (1983) talks about personal distress, which is related to one’s anxiety due to tensions in interpersonal contexts, and about fantasy, which is the use of fictional situations as a background to assume the character’s perspectives and feelings, although the latter has been the target of critics and reconsiderations (Koller et al., 2001).
The theme of sex differences regarding empathy has been frequently studied in psychology (Calandri et al., 2019; Di Tella et al., 2020; Eisenberg & Lennon, 1983). It is suggested that the idea that women tend to be more empathetic than men is based on social behavior stereotypes, which allude to the notion that female individuals are more caring and interpersonally oriented than males (Eisenberg & Lennon, 1983). Differences in cognitive tasks’ performances between men and women have been widely reported in the literature; consequently, even though women repeatedly perform better in those tasks, there is no consensus in the literature regarding sex divergences, even in research that specifically approaches the construct of empathy (Di Tella et al., 2020).
Sex differences are often evaluated by comparing men’s and women’s responses to self-report measures (Dryburgh & Vachon, 2019; Zhao et al., 2019). Although this kind of instrument has its advantages, it is strongly influenced by social aspects, such as social norms and expectations, for example (Dang et al., 2020). Self-report scales are also known to be affected by biased responses from participants who want to be considered socially adjusted and accepted by the community; this aspect that influences self-report instruments is called social desirability (Pavolov et al., 2021; Vesely & Klöckner, 2020). Accordingly, the use of self-report measure results regarding sex differences in empathy might be problematic, since the evaluated construct is characterized as a type of social ability, and, therefore, it usually comprises the desire for acceptance and social biases.
Self-report measures seem to evaluate how frequently a person acts and responds empathically to interpersonal situations, so they may diverge when measuring this construct between participants with different characteristics (Eisenberg & Lennon, 1983). Regarding sex differences, the differentiated levels of reported empathy by women may be related to a greater feminine disposition to perceive their behaviors and sentiments as represented by items that allude to empathic aspects, in comparison to men (Ventura-León et al., 2019), and it could influence the identification of participants with the stereotypes commonly attributed to their sex (Baez et al., 2017).
Multi-group confirmatory factorial analysis (MGCFA) has been used to verify if empathy is measured by self-report instruments in the same way for men and women. In this context, the invariance of empathy questionnaires related to the participants’ sex can be evaluated; if a measure is invariant, it means that group divergences are derived from individual differences, not methodological ones (Han et al., 2019). Whereas some recent studies point out the invariance of empathy scales (Grevenstein, 2020; Takamatsu et al., 2021), others attest that the instruments evaluate this social ability differently for men and women (Diotaiuti et al., 2021; Liu et al., 2018). Due to these diverse results in the literature, it is relevant to delineate an empirical study that seeks the identification of potential empathy measurements’ invariance regarding sex differences.
Self-report instruments continue to be often used to verify potential sex differences in empathy expression. Thus, it is common for research that evaluates empathy to utilize more than one self-report measurement, to present a more complex profile of empathic response (Azevedo et al., 2018; Zhao et al., 2019), especially by using the Interpersonal Reactivity Index (IRI) and the Empathy Quotient (EQ; de Lima & Osório, 2021; Zhao et al., 2019). This study’s objective consists of evaluating whether there are any divergences in the parameters of the cited empathy instruments concerning the participants’ sex by conducting MGCFAs and, therefore, concluding if those self-report measures are indicated to evaluate empathy in both men and women.
Method
Participants
Four hundred and seventy-seven adult participants were recruited. Participants’ ages varied from 18 to 59 years old (M=31.28, SD=11.34), and three hundred and sixteen were women (66.2%). The final sample was majorly composed of individuals from the state of Rio de Janeiro (72.1%), with an undergraduate degree (55.8%), and characterized as being single (59.3%).
Instruments
Two instruments were used to measure empathy. The Brazilian version of the Interpersonal Reactivity Index (IRI; Davis, 1980) was applied to evaluate empathy (Escala Multidimensional de Reatividade Interpessoal; Koller et al., 2001), as it measures the social ability through the reactions that participants report while observing experiences lived by others. It is formed of 17 items and three subscales: Empathic Concern (7 items; α=0.67), Perspective Taking (5 items; α=0.63), and Personal Distress (5 items; α=0.54). The adapted version to the Brazilian context did not include the Fantasy dimension, as the original one did (Koller et al., 2001).
The Empathy Quotient (EQ; Baron-Cohen & Wheelwright, 2004) was translated into Brazilian Portuguese (Quociente de Empatia) and adapted into a short version consisting of 21 items (Castelhano-Souza et al., 2018). It was composed of four dimensions: Cognitive Empathy (5 items; α=0.78), Emotional Reactivity (5 items; α=0.73), Social Abilities (5 items; α=0.73), and Emphatic Difficulties (6 items; α=0.57). The assessment of empathy was administered through the ratings of responses on a 4-point Likert scale.
The decision of choosing to focus on verifying if the IRI and the EQ are invariant when accounting for the sex variable stems from their frequent use in literature (de Lima & Osório, 2021). Since the two instruments were adapted to the Brazilian context in two different time periods, as the IRI was translated in 2001 and the EQ was validated in 2018, it would be interesting to attest whether these distinct scales may have significant differences in the assessment of empathy.
Procedures
Data were collected online and participants were recruited through invitations on emails and social media. The sample was assembled by convenience. All the participants had previously signed an Informed Consent Form. This study was performed after the approval from the Ethics in Research Committee.
Statistical Analysis
MGCFAs were performed to assess whether there were any differences in the empathy instruments (IRI and EQ) parameters regarding the participants’ sex. According to Roover (2022), the minimum sample size needed to adequately conduct the MGCFAs is equal to 300 individuals. The Robust Diagonally Weighted Least Squares (RDWLS) estimation method was used for categorical data (Li, 2016).
Many models have to be tested before measurement invariance can be supported for an instrument. However, since psychology scales tend to perform badly in some models due to the nature of the constructs they assess, it is widely accepted that the invariance can be evaluated through three models: configural, metric, and residual invariance (Counsell et al., 2020). The first model verifies whether the factors and number of items per factor, that is if the instrument’s structure is acceptable for both groups being tested. The second model analyses if the factor loads are equally significant for the participants’ groups. The final model evaluates the level of latent trait that is necessary to ensure scores’ categories (Cheung & Rensvold, 2002).
Goodness-of-fit indexes used to evaluate the models were: Root Mean Square Error of Approximation (RMSEA; values should be lower than 0.08 or, preferably, 0.06), Standardized Root Mean Square Residual (SRMR; values should be lower than 0.08), Comparative Fit Index (CFI; values should be higher than 0.90 or, preferably, 0.95), and Tucker-Lewis Index (TLI; values should be above 0.90 or, preferably, 0.95; Brown, 2015). Instruments’ invariances were assessed through the use of the CFI difference test (ΔCFI), which implies that, while analyzing a new model, if there is a reduction of CFI index levels of 0.01 and above, it is considered that the measurement invariance cannot be accepted (Cheung & Rensvold, 2002; Counsell et al., 2020).
Results
MGCFA results for the EQ can be observed in Table 1. Goodness-of-fit indexes allude to configural, metric, and residual invariance, thus demonstrating that the instrument is equivalent for both men and women. Therefore, it allows that group comparison to be performed.
Table 1: Multi-Group Confirmatory Factor Analysis (MGCFA) for the Empathy Quotient (EQ)
| Measurement Invariance | Goodness-of-Fit Indexes | ||||
|---|---|---|---|---|---|
| RMSEA [95% CI] | SRMR | TLI | CFI | ?#61508;CFI | |
| Configural Invariance Metric Invariance Scalar Invariance | 0.031 [0.019 - 0.040] 0.034 [0.025 - 0.043] 0.034 [0.024 - 0.043] | 0.061 0.064 0.062 | 0.988 0.985 0.985 | 0.990 0.986 0.986 | - -0.004 0 |
IRI invariance analysis can be seen in Table 2, and it showed that the scale did not support metric invariance. Consequently, the instrument is not equivalent for men and women and the comparison between groups is not recommended.
Table 2: Multi-Group Confirmatory Factor Analysis (MGCFA) for the Interpersonal Reactivity Index (IRI)
| Measurement Invariance | Goodness-of-Fit Indexes | ||||
|---|---|---|---|---|---|
| RMSEA [95% CI] | SRMR | TLI | CFI | ?#61508;CFI | |
| Configural Invariance Metric Invariance Scalar Invariance | 0.050 [0.043 - 0.058] 0.053 [0.045 - 0.060] 0.052 [0.045 - 0.059] | 0.078 0.082 0.079 | 0.920 0.912 0.914 | 0.929 0.919 0.917 | - -0.010 -0.002 |
After observing that metric invariance was not supported for the IRI, extra analyses were conducted to specify which items did not show equivalent factor loads between the two groups. Items’ metric invariance was tested by using the chi-square difference test, in which items should present an associated non-significant p-value (Byrne, 2009; Counsell et al., 2020). The results can be seen in Table 3. The use of this method requires that some constraints are imposed across groups (Byrne, 2009) and, because of that, the first items of each subscale (items 1, 2, and 4) were constrained and are not shown in the table below. Six of the instrument’s items were not invariant, from which four composed the Empathic Concern subscale and two were a part of the Personal Distress subscale of the IRI (see Table 4).
Table 3: Invariance Models’ p-values, Assuming that the Configural Model is Indeed Invariant
| Factor loads’ model (IRI item) | χ2 | p-value |
|---|---|---|
| Item 3 | 4.36 | 0.037* |
| Item 5 | 1.56 | 0.21 |
| Item 6 | 2.05 | 0.15 |
| Item 7 | 0.034 | 0.86 |
| Item 8 | 0.097 | 0.76 |
| Item 9 | 3.03 | 0.082 |
| Item 10 | 4.37 | 0.036* |
| Item 11 | 0.12 | 0.73 |
| Item 12 | 1.64 | 0.20 |
| Item 13 | 4.09 | 0.043* |
| Item 14 | 4.67 | 0.031* |
| Item 15 | 1.63 | 0.20 |
| Item 16 | 2.28 | 0.13 |
| Item 17 | 5.90 | 0.015* |
| Item 18 | 7.18 | 0.007** |
| Item 19 | 0.021 | 0.88 |
| Item 20 | 1.10 | 0.29 |
| Item 21 | 0.22 | 0.64 |
* = p < .05; ** = p < .01.
4: IRI Items that Did Not Present Invariance for Men and Women.
| Item | Dimension Evaluated | Item Description (version in Brazilian Portuguese) |
|---|---|---|
| IRI 3 | Empathic Concern | Sometimes I don't feel very sorry for other people when they are having problems. (Às vezes, eu não lamento por outras pessoas que estão tendo problemas). |
| IRI 10 | Empathic Concern | Other people's misfortunes do not usually disturb me a great deal. (As desgraças e problemas dos outros em geral não me perturbam muito). |
| IRI 13 | Empathic Concern | When I see someone being treated unfairly, I sometimes don't feel very much pity for them. (Quando eu vejo alguém sendo injustiçado, eu às vezes não sinto muita pena dele). |
| IRI 14 | Personal Distress | I am usually pretty effective in dealing with emergencies. (Geralmente eu sou muito efetivo para lidar com emergências). |
| IRI 17 | Empathic Concern | I would describe myself as a pretty soft-hearted person. (Eu descreveria a mim mesmo como uma pessoa de coração mole). |
| IRI 18 | Personal Distress | I tend to lose control during emergencies. (Eu tendo a perder o controle durante emergências). |
Discussion
As to discuss women’s tendency to report higher empathy levels in self-report measures, MGFCAs were proposed to test whether two empathy instruments could be used to test sex differences. Findings of instrument invariance suggest that the scale is adequate to evaluate a construct in the same matter in diverse groups (Counsell et al., 2020). EQ’s goodness-of-fit indexes showed that the measurement was invariant for the configural, metric, and scalar models, which allows comparisons between the tested groups (Han et al., 2019). Although not many studies have verified if the EQ is equivalent for men and women, sex invariance was attested in an Italian sample (Paolo Senese et al., 2018) and in an English sample (Allison et al., 2011).
While the EQ was found to successfully assess sex differences in empathy, IRI did not exhibit invariance for the metric model. Six items’ factor loads were not invariant between the two analyzed groups, from which four items referred to the Empathic Concern subscale and two were attributed to the Personal Distress subscale.
Among the most recent research regarding IRI’s sex invariance, the instrument was reported as equivalent for men and women in Spanish (Lucas-Molina et al., 2017), and German (Grevenstein, 2020) adult samples. It is noteworthy that while the Chinese study’s adapted version of the IRI was invariant, it had a different factor structure from the original scale: whereas Davis (1983) stipulated four dimensions for the instrument, the Chinese authors presented a bifactor model that highlighted the Empathic Concern and Perspective Taking dimensions (Wang et al., 2020). In contrast, the measurement’s lack of invariance for the scalar model was attested in a sample of Colombian ex-combatants (Garcia-Barrera et al., 2017).
The IRI version used in the present study did not follow the original four dimensions proposed by the author of the original scale (Davis, 1980). Many researchers have classified the IRI’s original factor structure as inadequate throughout the years and have also proposed other structural configurations (Garcia-Barrera et al., 2017; Lucas-Molina et al., 2017; Murphy et al., 2020; Wang et al., 2020). Since factor structure problems have influenced other studies, the present research’s instrument configuration might not be the best fit and, because of that, it might have affected the MGCFA. Moreover, it is also worth mentioning that the Brazilian variant of the IRI was adapted and validated in 2001 (Koller et al., 2001). A recent systematic review and meta-analysis verified that only two studies assessed the IRI’s test-retest reliability (de Lima & Osório, 2021), so there is not enough evidence to stipulate whether the scale can be used to measure empathy in the same manner after some time has passed.
Although the IRI’s factor structure and temporal problems might have played a part in the lack of equivalence between groups, other aspects might have also influenced it. Most of the non-invariant items were a part of the Empathic Concern subscale. This affective empathy component involves genuine feelings toward others (Davis, 1983), especially worry (Koller et al., 2001). While empathy’s cognitive dimension is often not affected by sex differences, the affective aspect presents more differential results, as it is more influenced by gender-specific expectations (Kamas & Preston, 2021). Accordingly, groups’ non-invariance might be a consequence of participants’ gender differences, instead of sex-related ones. To act correspondingly to society’s expectations, men could report lower levels of empathic concern in a way to not appear or to not be considered too feminine (Kamas & Preston, 2021).
The idea that gender-specific social expectations might have contributed to the lack of invariance can also be verified when the Empathic Concern subscale’s items are analyzed. Item 17 (“I would describe myself as a pretty soft-hearted person”) involves the term “soft-hearted”, which may elucidate different meanings for men and women. Research conducted in over fifty countries, including Brazil, described that this adjective was considered by participants as being a feminine personality trait (Williams & Bennett, 1975). More recently, studies executed in Jamaica and Barbados (Walters & Capertener, 2017), and Spain (Castillo-Mayén & Monten-Berges, 2017) also reported that both men and women associate the term with a more feminine characteristic. Thus, the item’s invariance may be related to social expectations and pressure on men not to act similarly to women.
The other three non-equivalent items of the Empathic Concern subscale (items 3, 10, and 13) may also have been influenced by social factors, especially by social desirability. The items’ statements, which include feeling pity, disturbed, or worry towards others, might act as a trigger for conscious or unconscious distortions of responses (Hauck Filho & Valentini, 2019). Men and women might be differently affected by the impact of having to choose the most sociably accepted answer and which one that could be.
The three aforementioned items are also characterized by being the only reverse items of the Empathic Concern subscale. The psychometrical properties of an instrument are often hindered when the measurement contains negative items since participants have more difficulty in assessing the item’s meaning when it’s inverted (Kam et al., 2021; Vigil-Colet et al., 2020). Consequently, individuals do not always understand a construct or an item the same way and, because of that, group comparison is also undermined when the instruments comprise inverse items (Suárez-Álvarez et al., 2018). Therefore, the fact that the previously mentioned items were constructed inversely might have contributed to the differentiated measure of empathic concern by men and women in this study.
Two items of the Personal Distress subscale were not invariant in this study. Personal distress is distinguished from empathic concern because the feelings of worry experienced by an individual are directed toward the self, not others, in an emotionally demanding context (Israelashvili et al., 2020). Nevertheless, this study’s two non-invariant items of the Personal Distress subscale do not include themes or statements that make emotional defying conditions explicit or implicit. The term “emergencies” might be too general and broad for Brazilian participants’ full comprehension of the construct since it might not allude to the emotional response that those items should generate. Conceptual equivalence is an important aspect to be examined while a psychological measurement is translated and adapted to different languages as it evaluates if a specific term or expression, even if it is adequately translated, measures the same construct in diverse cultures (Khouri & Silva, 2019). Those items may not be suitable to represent personal distress and, accordingly, the participants may not interpret them the same way. That could have contributed to the lack of metric equivalence between men and women.
Item 14 of the Personal Distress subscale (“I am usually pretty effective in dealing with emergencies”), in addition, was classified as troublesome in the step of the statistical validity of the instrument. The aforementioned item might not be an adequate measure of the construct, thus it can be removed from the Brazilian version of the IRI (Koller et al., 2001). Besides the conceptual equivalence problem presented by the item, it did not show the necessary psychometrical properties.
Measurements that do not display group invariance do not necessarily measure different constructs, but they might have diverse meanings attributed to each of the groups (Han et al., 2019). As a result, the absence of metric invariance presented by the IRI might mainly reflect a masculine difficulty in expressing worry about others, which is most possibly related to the fear of not behaving as it is expected by the respective gender and the influences that men suffer throughout their childhood and adolescence. Future research should not only analyze empathy’s sex differences but also investigate gender’s influence on empathy. Moreover, it should be noted that problems associated with items’ writing (e.g., social desirability and terminology issues) and psychometrical (e.g., an item that did not contribute to the scale’s confiability), and structural properties (e.g., reverse items) of the Brazilian version of the IRI might have also influenced the lack of metric invariance for some items.
Conclusions
This study’s objective was to analyze two empathy measurements to find out if there were any measurable differences in empathy for a sample of Brazilian adult men and women. In agreement with this theme’s literature, EQ is shown to be an invariant instrument and, thus, can be used to compare men’s and women’s perceptions of their empathy.
In contrast, one should be careful when evaluating empathic concern’s sex differences, since the IRI was not considered to be equivalent for this study’s sample. Those non-invariant IRI items may be influenced by gender and social expectations. In addition, social desirability, psychometrical, temporal, and adaptation issues might have also contributed to the instrument’s aforementioned results. Future studies may still use the IRI to measure empathy, but they will need to take into consideration that some items seem to be problematic when comparing men and women.
One of the present study’s limitations refers to the administration of only two empathy instruments to test their invariance. Since empathy is a complex construct, the use of other forms of measure might provide more information and enhance its comprehension, especially because empathy involves many different theoretical frameworks.
This study contributes to the research of empathy’s sex differences by being the first in Brazil to test empathy instruments’ invariance. The results reported here may highlight the importance of verifying if the psychological measurements used by research are adequate to measure what is expected. Therefore, the present study’s results may be able to help researchers decide which scale is the most suitable to be used in the Brazilian context, outlining that the more recent instrument seems to be more robust when considering the sex variable.
Whereas this study has reiterated the existence of a sex difference concerning empathy, it has also pointed out that not all self-report empathy instruments are appropriate to measure such difference, as men and women seem to comprehend and answer some items (especially those that comprise empathy’s affective components, such as empathic concern) differently. Consequently, men’s and women’s divergence regarding empathy seems to influence empathy measurements.
