Open Access Repositorium für Messinstrumente


Emotional Empathic Drive Short Scale (EED)

  • Autor/in: Karlstetter, W.
  • In ZIS seit: 2017
  • DOI:
  • Abstract: Empathy is a fundamental precondition for human interaction, since it enables us to anticipate the intentions of others via sharing of and thinking about their feelings – if we want to. Using items fr ... mehrom established empathy questionnaires, this study provides the first questionnaire of empathy which is valid and reliable, and yet short. The Emotional Empathic Drive Short Scale (EED) is tied to a comprehensive theoretical framework of motivated empathy (Zaki, 2014) and measures the dispositional drive to share others’ emotions. It has been developed with samples from Canada and Germany, and is available in English and German. The EED short scale resembles theoretical expectations regarding its associations with other constructs such as gender, authoritarianism, and social dominance orientation. weniger
  • Sprache Dokumentation: English
  • Sprache Items: englisch, deutsch
  • Anzahl der Items: 5
  • Reliabilität: McDonald’s omega = .70; retest reliability = .62
  • Validität: evidence for construct validity and criterion validity
  • Konstrukt: empathy, emotional empathic drive
  • Schlagwörter: empathy, identification
  • Entwicklungsstand: validated
    • Instruction

      Before the five items of the Empathic Drive Short Scale (EED) are presented, respondents read: “Below is a list of statements. Please read each statement carefully and rate how frequently you feel or act in the manner described. There are no right or wrong answers or trick questions. Please answer each question as honestly as you can.” The response categories for each of the five items are (0) never, (1) rarely, (2) sometimes, (3) often, (4) always.



      Table 1

      Items of Empathic Drive Short Scale






      It upsets me to see someone treated disrespectfully.


      Mehrabian, 2000


      I remain unaffected when someone close to me is happy.


      Hornak et al.1996; Mehrabian, 2000


      I have tender, concerned feelings for people less fortunate than me.


      Davis, 1983


      I become irritated when someone cries.


      Spreng et al. 2009


      I am not really interested in how other people feel.


      Spreng et al. 2009


      Response specifications

      Unlike most available empathy questionnaires, the proposed response format based on Spreng, McKinnon, Mar, and Levine (2009) asks respondents to report how frequently they feel or act in the manner described. Marcus, MacKuen, Wolak, and Keele (2006) point out that in the case of items intended to observe emotions, asking for either the frequency of emotional experiences or their intensity preserve the multidimensional structure of a scale better than agree-disagree scales. A survey experiment of empathy items (including the items of the EED short scale) revealed that the common response format which asks respondents how much the item statement describes them leads to significantly higher reported values (Karlstetter, 2014) than the response scale with frequencies. Most likely, this is due to increased social desirability ‘encouragement’. Therefore, the response format with frequencies paired with the instructions above has been chosen for this short scale.



      The five items of the EED scale can be added up to form sum scores for further correlational analyses, because the scale shows essential tau-equivalence (see item analyses). Hence individual item weights are not required. Accordingly, the EED sum score can assume values between 0 and 20. In fact, almost the full range of the scale has been used by respondents (see table 2).


      Table 2

      Descriptive statistics of the EED scale (sample 3, n = 277)



      Standard Deviation













      Application field

      The EED can be used to observe empathy and/or cross-validate other measurement instruments of empathy in social science, psychology, and neuroscience. The items forming this scale have been applied for decades in numerous studies and surveys on empathy (see below). The particular item compilation at hand has been tested with self-administered printed questionnaires. Native speakers of English or German of age 14 and older are the target group of the EED.

    From a neuroscience perspective, empathy is a fundamental human trait that can be found across cultures (Walter, 2012). Empathy ‘carries’ various emotions of others into our minds and bodies so that we can experience (or at least anticipate) what others feel - and thus are able to comprehend their situation and their intentions. It is fundamental, since it allows us to interact and cooperate with one another –capabilities which have contributed greatly to our species’ success (see Zaki & Ochsner, 2012). Despite various conceptualizations and measurement techniques (for an overview see Cuff, Brown, Taylor, & Howat, 2016), scholars in psychology and social cognitive neuroscience agree that there are at least two main dimensions of empathy. Cognitive empathy means mentalizing the emotions of others, whereas affective empathy means sharing the emotions of others(see Batson, 1991; Hoffman, 2000; Decety & Jackson, 2006; Jolliffe & Farrington, 2006).

    Besides the emotional and cognitive dimensions, there are two different ways in which empathic responses to external stimuli can be activated (see Vogeley, Schilbach, & Newen, 2013). First, empathy can be reflexive and automatically activated. Intention or any conscious or unconscious motivation is not involved. This view has become dominant across multiple disciplines including developmental psychology (Hoffman 1982; Eisenberg, 1989), ethnology (Preston & de Waal, 2002; de Waal, 2010), social psychology (Hatfield, Cacioppo, & Rapson 1994; Hatfield, Forbes, & Rapson 2013), and neuroscience (Gallese, 2007; Decety, 2010). Corresponding models help to explain social phenomena such as the spread of laughter (Provine & Young, 1991), panic (Kerckhoff & Back, 1968), and yawning (Rundle, Vaughn, & Stanford 2015). Second, there are empathic processes which do not happen involuntarily but are shaped fundamentally by context factors. These factors include observer-target similarity, how much the observer values the target, perceived power differences, and perceived need (see Batson, Eklund, Chermok, Hoyt, & Ortiz 2007; Eklund, Andersson-Stråberg, & Hansen 2009; Galinsky et al., 2006), as well as political boundaries (Mitchell, Macrae, & Banaji 2006), or any kind of social category by which others are believed to be different from one’s in-group (see Harris & Fiske, 2006; Hein, Silani, Preuschoff, Batson, & Singer  2010).

    Borrowing from theories on information processing (Kunda, 1990; Taylor & Brown, 1988; Taber & Lodge, 2006) and emotion regulation (Lazarus & Folkman, 1984; Gross, 2002; Coan, 2011), Zaki (2014) accounts for the ample evidence of a motivated activation of empathy by providing a comprehensive theoretical framework. This framework builds the theoretical background of the empathy short scale at hand. It explains under which circumstances different types of motives stimulate individuals to regulate empathy up or down. The model identifies emotional and cognitive empathy, as well as mind perception as a third dimension of empathy. The latter dimension means that individuals need to detect that another has some internal emotional state before they can attempt to share or understand this state.

    Zaki (2014) describes three motives stimulating individuals to engage in empathy and three motives causing people to avoid engagement in empathy and the emotions that come with it. These motives shape all three dimensions of empathy in social context. Perhaps the most intuitive motive is that individuals want to avoid (psychological) pain. That is, another’s negative emotion will put psychological burden on the observer, if she shares this emotion. Individuals might also want to avoid material cost. As empathy is stimulating potentially costly helping behavior such as donating money, individuals apprehending this are motivated to avoid such situations altogether (see Pancer, McMullen, Kabatoff, Johnson, & Pond, 1979). The third motive to avoid using one’s empathic capacities is its potential interference with competition. In situations in which people are competing with others or have to negotiate with hostile others, they seek to regulate empathy for these others down (Galinsky, Maddux, Gilin, & White, 2008; Cikara & Paluck, 2013).

    In contrast, individuals are motivated to share others’ feelings for several reasons. First, humans want to share positive emotions of others (Morelli, Lieberman, & Zaki 2015) – just as they are motivated to avoid negative emotions (see above). For example, observes help others especially when they see a chance to share the positive emotions stimulated in the individuals they help (Zaki, 2014; also see Smith, Keating, & Stotland, 1989). Second, individuals are motivated to share the emotions of others in order to strengthen or maintain social bonds. For example, individuals who feel rejected by others and have a greater need to close social distance to others attend more to emotions of others and interpret their social cues more correctly (e.g. Kraus, Cote, & Keltner, 2010). Finally, empathy is applied since it is desirable. There are two kinds of desirability which motivate empathy. For individuals who have a self-image and hold values which are consistent with applying empathy or emotional outcomes such as compassion, using empathy reaffirms one of one’s self-image (Zaki, 2014; see Frankl, 1985). However, even if individuals do not ‘believe’ in empathy itself, they might very well be motivated to use it in order to appear socially conform (i.e. social desirability). For example, empathy is more socially desirable among women than men, leading to consistently higher scores of women as compared to men on empathy questionnaires as well as behavioral tasks (e.g. Davis, 1983; Ickes, Gesn, & Graham, 2000).

    In real social situations multiple of these motives can be apparent. If empathy is ultimately engaged in or not will be dependent on an individual’s dispositional drive to engage in empathy and the ‘net account’ of approach versus avoidance motives apparent in a given situation. For example, when higher social desirability of empathy among women is compensated, such as through monetary incentives (Klein & Hodges, 2001) or by making heterosexual men believe that women find social sensitivity attractive (Thomas & Maio, 2008), then men will use empathy just as much and with similar accuracy than women.


    What this short scale measures

    The items forming the emotional empathic drive short scale (see tab. 1) have been originally constructed by different authors and historically before a comprehensive theory of empathy has been developed (see scale development). Therefore, I reinterpret what this particular item compilation measures relying on Zaki’s (2014) theoretical framework of motivated empathy. While Zaki distinguishes three dimensions of empathy (cognitive empathy, emotional empathy, mind perception) and focuses much on situations motivating the avoidance or engagement in empathy, the Emotional Empathic Drive scale at hand reflects an unidimensional understanding of empathy as an individual disposition – to be more or less motivated to use empathy across situations, and to therefore tend more or less to share others’ emotional states.

    More precisely, EED is the drive to overcome motivations which otherwise would lead one to not use one’s capabilities to experience others’ affective states. Except item 5, all items of the short scale stimulate the respondent to imagine oneself in a social situation. The situations potentially involve motivations either to engage in empathy and to regulate one’s emotions up (item 2) or to avoid empathy and to regulate one’s emotions down (item 1, 3 and 4). For example, seeing “when someone cries” (item 4) potentially leads to distress through emotion sharing, which individuals are motivated to avoid in some way (e.g. through avoiding the situation or reappraisal). The scale has a focus on respondents’ drive to share the negative emotions of others (i.e. to overcome an avoidance motive), while item 2 taps into individuals’ drive to share positive emotions of others (i.e. to pursue an approach motive). Item 5 is a general assessment of EED which does not stimulate respondents to imagine a particular social situation. More precisely, it taps into the process prior to but closely related to affective sharing, namely mind perception: An individual who tends to disregard emotional states of others in the first place, is not very likely to share their emotions either. As outlined further in the section about scale development, empirical evidence also implies the one-dimensionality of items 1-4 capturing affective sharing and item 5 capturing mind perception. In a broader sense, with the EED short scale a respondent will report his/her drive to decrease social distance, even if that causes one to experience negative emotions. Higher empathy values indicate higher levels of individuals’ drive to decrease social distance (i.e. to reach out to others) despite of negative emotions transmitted from the other to the self or because others experience some positive emotion which one can share only through empathic engagement. Hence, across social situations, an individual with a higher EED score is more likely to be relatively more triggered by motives to share others’ emotions than by motives not to do so.

    That no short scale of empathy exists as of today (to my best knowledge) might be due to previous empathy questionnaires having a psychological/psychiatric purpose and/or being therapeutically used (e.g. Baron-Cohen & Wheelwright, 2004); Davis, 1983; also see Chlopan, McCain, Carbonell, & Hagen, 1985). Thus, due to a lack of an adequate – valid, reliable, and practical/short – questionnaire to measure empathy, this construct has been rarely included in broader social science studies, for example, to explain political attitudes (but see Newman, Hartman, Lown, & Feldman, 2015). In order to observe this fundamental human trait in research settings where multiple questionnaires are combined and multiple constructs are observed, such as in large scale (cross-country) surveys, a short questionnaire of empathy consisting of ‘thoroughly picked’ items is needed.


    Item generation and selection

    A four-step approach led to the final EED short scale. The first step was an extensive and interdisciplinary literature study of the phenomenon of empathy, including a review of available (full length) empathy questionnaires (see Karlstetter, 2015, 2014). About a dozen questionnaires have been developed between 1949 (Dymond, 1949) and 2011 (Reniers, Corcoran, Drake, Shryane, & Völlm, 2011) to observe empathy. In general, recently constructed questionnaires are greatly relying on items from previous questionnaire instruments, using them in different item combinations and reflecting different focuses and understandings of empathy (see Hogan, 1969; Mehrabian & Epstein, 1972; Davis, 1983; Baron-Cohen & Wheelwright, 2004; Jollifee, D. & Farrington, 2006; Reniers et al., 2011; and others). Most empathy questionnaires attempt to measure between one and three dimensions: sharing emotions, understanding emotions, and some sort of (pro)-social motivation and/or ability. A review of the neuroscience literature on empathy reveals that even the ‘big’ and presumably separate concepts affective empathy and cognitive empathy are highly interrelated and commonly coactive in social situations (Zaki & Ochsner, 2012). This view is in accordance with survey studies which find empathy items mostly falling on one or two correlated sub-dimensions (affective and cognitive) empirically, although they have been theoretically split into more facets (see Karlstetter 2015, 2014 for a discussion).

    Reflecting a variety of theoretical frameworks (see Cuff et al., 2016 for a comprehensive overview) there are multiple questionnaires and hundreds of items available from which one can choose. In light of this challenging theoretical and methodological situation Spreng et al. (2009) took a different approach. Compiling 142 empathy items from previous questionnaires, their goal was to find a common factor which these questionnaires/items share – reflecting a ‘minimum consensus’ of the previous literature on empathy. Numerous empathy items, which are linked to different theoretical understandings of empathy, have been validated, sorted out, and finally combined in the Toronto Empathy Questionnaire (TEQ; Spreng et al., 2009). Unlike other empathy questionnaire instruments, TEQ items have been validated through behavioral tests as well. The “Reading the Mind in the Eyes Test” (MIE; Baron-Cohen et al., 2001) and the “Interpersonal Perception Task-15“-Test (IPT-15) (Archer, Costanzo, & University of California System,1994; Sagi & Hoffman, 1976; Ungerer, 1990; Zahn-Waxler, Friedman, & Cummings, 1983) have been applied. By using a uni-factorial explorative approach (EFA), the authors assembled 16 items in the TEQ. It consists of items from the Balanced Emotional Empathy Scale (BEES; Mehrabian, 2000), the IRI (Davis, 1983), and the HES (Hogan, 1969). The 16 items assess one common factor, namely emotional empathy (Spreng et al., 2009). The TEQ is much shorter than other empathy questionnaires, which contain usually about twice as many items, while still showing enhanced psychometric properties as compared to previous questionnaires (Lietz, Gerdes, Sun, Geiger, Wagaman, & Segal, 2011). Nevertheless, the TEQ is still too long for measuring empathy in large-scale (representative) social science studies.

    Using the TEQ as a starting point, the second step was a systematic scale reduction process based on Stanton, Sinar, Balzer, and Smith (2002). Considering external, internal, and judgmental item criteria, the best performing items of the TEQ were identified (see Karlstetter, 2014). Every TEQ item was correlated with specific indicators/criteria. The external constructs that have been applied in scale development, for example, the cognitive empathy sub-scale “perspective taking” from the Interpersonal Reactivity Index (IRI; Davis, 1983; see above), right-wing-authoritarianism (Altemeyer, 1981), and gender, test the scale’s general setting in relation to theoretically ‘close’ and ‘far’ external measurements (i.e. convergent and discriminant validity; for all validation constructs see quality criteria). Then, these correlations were transformed into item quality scores. A total external validity sum score was constructed from the scores for every item in order to enable meaningful comparisons between items. In addition to statistical analyses, analysis of judgmental item validity has been conducted, which focuses on theoretical and semantic considerations such as the ambiguity and semantic redundancy of items (see Stanton et al., 2002). Furthermore, I incorporated respondents’ survey pre-test comments into the analysis of item wordings (pre-test for sample 2; Karlstetter, 2014). Based on external and internal scores (i.e. item-remainder correlations) as well as judgmental analysis, the most ‘promising’ items were identified.

    The third step in the scale development was a dynamic combination of theory testing and optimizing the short scale’s item configuration by applying (multigroup) CFA (across genders as well as samples from Canada and Germany; see Karlstetter, 2015, 2014). This led to a reduced TEQ of about one fourth of its former length. The item reduction process revealed that the TEQ items 3, 4, 6, 11, 12, and 13 are the best performing items considering all applied external, internal, and judgmental criteria. Except item 13, the EED scale at hand consists of these items out of the full-length TEQ. All empirical analyses have been conducted using STATA 14.

    This new short scale reflects a balanced item selection to measure empathy. It is based on the premise that for an individual item to be selected for the final item compilation, it is less important how much it correlates with a specific construct (for instance perspective taking) – merely picking an item that correlates the highest. Instead, it is more important how those items which reflect different association 'focuses' are combined, in order to cover empathy in the best (quantitative and qualitative) way (see Stanton et al., 2002). More precisely, one assumption underlying this procedure is that items with similar correlation patterns are redundant. In addition, items might illustrate a complementary relationship to each other (concerning correlations with various criteria). As a result, fewer items might constitute similar performance than more. The third assumption and analytical focus, respectively, is that items which perform exceptionally high concerning certain criteria – and like no other item – should remain in the short scale: since these items are unique in some way, their structural patterns cannot be replaced by other items. Finally, this quantitative approach has been accompanied with theoretical considerations. Most importantly, the short scale consists of items which are coherent with a motivated understanding of empathy (in contrast to conceptualizing empathy as a ‘pure’ ability).




    Table 3

    Sample Characteristics


    Sample 1

    (Spreng et al., 2009)

    Sample 2

    Sample 3

    Samples 4 & 5 (Spreng et al., 2009)





    79 & 66









    Paper (assisted by interviewer)


    Country (Language)


    Germany (German)

    Germany (German)

    Canada (English)


    Exploratory Factor Analysis (EFA), Confirmative Factor Analysis (CFA), and Multigroup CFA of the full 16-item Toronto Empathy Questionnaire (TEQ).

    EFA and Multigroup CFA of full TEQ; Correlations with multiple constructs.

    Non-student sample for EFA and Multigroup CFA of reduced TEQ; Correlations with multiple constructs.

    Correlations of full TEQ with behavioral tests in sample 4;

    Re-Test Correlations between both TEQ samples.

    % of Women





    Age [M (SD); range]

    18.82 (1.22);

    20.85 (1.84);

    35.92 (16.75);


    18.90 (2.99);



    College Students

    College Students

    Below Middle School: 2.11%

    Middle School: 24.91%

    High School: 45.96%

    University Degree: 22.46%

    Other: 1.05%

    College Students

    Notes: Education levels in sample 3 have been translated from German into English: “Kein Schulabschluss” & “polytechnische Oberschule nach 8.Klasse” & “Volksschulabschluss” = “Below Middle School”; “Hauptschule” & “10. Klasse der polytechnischen Oberschule” & “Realschulabschluss, mittlere Reife” = “Middle School”; „Abitur, Fachhochschulreife, 12. Klasse der polytechnischen Oberschule“ = “High School”; „abgeschlossenes Studium an einer Hochschule / Fachhochschule“ = „University Degree“; and „anderer Abschluss“ = „Other“.


    In order to control for question order effects, participants have been randomly assigned to one of two opposite questionnaire orders in each German sample. Students participating in the first German study applying the TEQ (sample 2; see tab. 3) have received no credit. For sample 3, students interviewed respondents on credit basis after interview training. Systematic sampling was applied to identify respondents. Interviewers used fixed intervals (e.g. every fourth person) in order to select respondents on the street. Each participant was handed a questionnaire and then filled it out anonymously. For further descriptions of samples 1, 4, and 5, please refer to Spreng et al. (2009).


    Item analyses

    Exploratory Factor Analysis

    The results indicate that the items of the EED short scale load on one common factor. The Eigenvalues drop considerably from the first to the second factor in each sample. In particular, the Eigenvalues of the first two factors are 1.26 and .24 (sample 1), 1.47 and .08 (sample 2), and 1.52 and -0.5 (sample 3), respectively. With the exception of item 1 in sample 2, all items load sufficiently high on this common factor. Kaiser-Meyer-Olkin (KMO) values have been computed in order to analyze the samples’ adequacy for use in an EFA. KMO values of .75, .69, and .79 (from sample 1 to sample 3), respectively, indicate that the variables have enough in common to warrant a factor analysis.


    Table 4

    Uni-dimensional Loadings of Explorative Factor Analysis in three Samples

    Item No.

    Sample 1
    (Canadian Student Sample, n = 200)

    Sample 2
    (German Student Sample, n = 260)

    Sample 3
    (German Systematic Sample, n = 276)





















    Notes:    a) Principal factor analysis was conducted using STATA 14.    b) Only cases with non-missing values on all variables have been included.


    Confirmatory Factor Analysis

    To further investigate both the dimensionality of the EED and how well each item is explained by the common factor, confirmatory factor analysis (CFA) was conducted (sample 3; n = 276; only cases with non-missing values on all variables have been included). First, a congeneric measurement model has been estimated in which the loading of item 1 was constrained to 1. Measurement errors were not allowed to correlate between items. Maximum likelihood (ML) was used as the estimator with observed information matrix for standard errors. Each item loads on the latent factor with at least .50 (see fig.1).Overall, the model fits the data very well: χ² = .90, df= 5, p = .970; CFI = 1.000; TLI = 1.041; RMSEA = .000; SRMR = .010. Using bootstrapping (100 samples) to estimate standard errors leads to very similar results.


    Figure 1. t-congeneric measurement model of emotional empathic drive showing standardized loadings and standardized measurement errors (n3 = 276).


    Second, an essentially tau equivalent model has been tested by setting the factor variance to 1 and constraining all item loadings to be equal. With otherwise identical estimation parameters, this model fits the data very well: χ² = 7.03, df= 9, p = .634; CFI = 1.000; TLI = 1.011; RMSEA = .000; SRMR = .034. This model produces similar standardized item loadings as in the unconstrained model. Comparing χ² and degrees of freedom between the two models (χ²(df) diff = 6.13(4)) shows that the increase in χ² in the tau-equivalent model is insignificant. The RMSEA, which allows for model comparison, is identical across models. Again, using bootstrapping (100 samples) to estimate standard errors leads to very similar results.

    Taken together, the CFA indicates that the items of the short scale observe one latent construct. Furthermore, the excellent fit of the essentially tau-equivalent model indicates that the sum scores of the EED scale can be used for further correlational analyses without the need to apply weights to the individual items.



    The EED short scale can be objectively applied by researchers as well as interviewers, since it is a standardized measurement instrument. More precisely, the instructions for participants, the questionnaire items, as well as the response format are standardized. The latter aspect in combination with the instructions for the researchers regarding the sum score calculation and scaling polarities ensures that responses on the EED scale are objectively evaluated as well. Furthermore, orientation on a clear and systematic scale reduction process has contributed to objectivity regarding decisions on which items to select for the EED short scale. Finally, the scale’s theoretical framework has been described transparently. As such, observed values of empathy using the EED scale can be objectively compared across researchers and studies.



    Although still widely used to prove a scale’s reliability, it has been demonstrated that Cronbach’s alpha lacks certain methodological qualities. One of the available measures which overcome disadvantages of the coefficient alpha is McDonald’s omega (McDonald, 1999; also see Raykov, 1997; Raykov & Marcoulides, 2011). Omega has been calculated for the systematic sample (sample 3) following the approach suggested by Dunn, Baguley, and Brunsden (2013) and applying the package MBESS in R (bootstrapping 10000 samples).The computed ω of .70 (95%-CI = .63; .77) indicates acceptable scale reliability. The lower bound of ω falls below .70, indicating that future (representative) survey studies should further investigate the short scale’s reliability. In addition, re-test reliability has been calculated by correlating the sum scores of the EED scale between the Canadian samples. 66 individuals responded both on the questionnaire for sample 4 and 57 to 84 days later on the questionnaire for sample 5 (on average 61.1 days later). The computed r of .62 indicates that the short scale is sufficiently reliable (see Aiken & Groth-Marnat, 2006).



    Content validity

    In general, it is difficult to prove content validity of a measurement instrument by empirical investigation. The EED scale consists of established items of previous empathy questionnaires, which have been re-validated and found to measure empathy by experts on empathy in numerous studies across various disciplines (see above). Furthermore, the remaining items of the EED scale are closely aligned with the theory of motivated empathy, as summarized by Zaki (2014). Hence, content validity of the EED scale is established.


    Construct validity

    The Interpersonal Reactivity Index (IRI: Davis, 1983) is frequently used as a validation instrument of empathy measures (e.g. Lietz et al., 2011; see Konrath & O’Brien, 2011). In the study at hand, two of its four subscales – Empathic Concern (EC) and Perspective Taking (PT) –, which are known to be more valid and robust than the remaining two IRI subscales (Reniers et al., 2011; Lietz et al., 2011), are used in addition to the full TEQ questionnaire in order to test convergent validity. Furthermore, as the individual tendency to see the world as a dangerous place in which authority and tradition ‘ease’ potential threats caused by others, right-wing authoritarianism (RWA; Altemeyer, 1981) was expected to correlate negatively with the EED scale (see Feldman & Johnston, 2014; Hopf, 1998; Rokeach, 1960; Alford, Funk, & Hibbing, 2005). The third validation construct is a general preference for intergroup hierarchy and social inequality, namely social dominance orientation (SDO; Sidanius & Pratto, 1999). In the motivated empathy framework, which underlies the EED short scale, Zaki (2014) explicitly describes SDO as an operationalization of the empathic motive to avoid interference (with competition; see theory). Indeed, empirical studies on SDO and empathy demonstrate that these constructs are highly negatively associated (Sidanius et al., 2013; Esses, Veenvliet, Hodson, & Mihic, 2008).

    Given that the short scale of emotional empathy (EED) consists of only five items, it replicates its corresponding full length scale, the TEQ, sufficiently in the German sample (see tab. 5, sample 3:r =.87). In the Canadian samples (samples 1 & 4), the correlations between the short scale and the TEQ are .86 and .82, respectively. More importantly, the EED scale replicates the TEQ’s associations to the IRI and its subscales PT and EC (see Spreng et al. 2009), indicating substantially stronger associations with emotional (r =.78; for German student sample 2: r = .73) than cognitive processes of empathy (r = .45; for German student sample 2: r = .32).

    As expected, EED is negatively associated with RWA. In order to measure RWA, I used a three-item short scale (in German) from Schmidt, Stephan, and Herrmann (1995, based on Lederer, 1983) including items such as “In general, it is good for a child to adapt to its parents' values and attitudes.” Using a 4-point Likert response scale, this short scale measures the core component of authoritarianism, obedience under authorities (see Schmidt & Heyder, 2000, p. 447).


    Table 5

    Validation of the EED short scale



    Full TEQ1


    Empathic Drive Short Scale2

    Full TEQ



    IRI: Perspective Taking



    IRI: Emotional Concern



    RWA: Authoritarian Obedience



    SDO: Group-based Dominance



    Prosocial Political Attitude








    .03 (p = .59)




    Notes: Pearson’s r has been calculated based on sum scores of corresponding constructs/measurements. Correlations are significant (p < 0.01) unless reported otherwise.1 Correlations based on sample 2 (n = 260) using the full 16-item questionnaire of the Toronto empathy questionnaire (TEQ). 2 Correlations based on sample 3 (n = 276). Only cases with non-missings on all variables have been used.


    The EED scale is negatively associated with SDO to a remarkably high extent. The three items used to assess SDO in this study focus on group-based dominance (see Zick, Wolf, Küpper, Davidov, Schmidt, & Heitmeyer, 2008). The scale includes items such as “Some groups of people are less worthy than others” to which individuals responded on a 4-point Likert scale.

    In order to test if the EED measures a stable drive to overcome context-dependent motivations not to share the negative emotions of others, sample 3 also included an item asking respondents “Government should make sure that war refugees from abroad get asylum in Germany”. The corresponding correlation (r = .36) is in line with the theoretical expectation. The wording does not include adjectives or nouns indicating emotional states of potentially affected refugees, such as “sad”, “hurt”, “depression” etc. Thus, respondents have to imagine (cognitive empathy) affective states of refugees on purpose and need to be motivated to share them. “War” in combination with “refugees” indicates to participants that negative emotions are likely to be experienced by others. If they are driven to overcome potential distress caused by that sharing, the goal to act politically is pursued and expressed explicitly. In addition to this cognitive and emotional burden, competition (i.e. potential ingroup-outgroup conflict over resources) is a fundamental motive to suppress state empathy, which is almost certainly apparent when respondents answer this question. Given that, the positive and mediocre correlation with EED indicates that the latter indeed measures a stable drive to overcome context-dependent motivations not to share the negative emotions of others.


    Criterion validity

    Many quantitative studies on self-report empathy found that women report significantly higher empathy levels than men (Schieman & Gundy, 2000; Davis, 1983; and others). I suggest that a short scale should reproduce this well-documented empirical phenomenon of self-report questionnaires, in order to be valid and reliable. In fact, the EED scale replicates previous survey studies with a correlation of −.32 (see tab. 5; r = −.37; p< .01 in sample 2).

    In contrast, a null relationship between empathy and age was expected (see Grühn, Rebucal, Diehl, Lumley, & Labouvie-Vief, 2008). More precisely, while (a drive for) perspective taking (i.e. cognitive empathy) might decrease somewhat with age; becoming more effortful as part of a general decrease in mental abilities, this is not expected for the drive to share others’ emotions. Accordingly, an insignificant correlation of .03 has been found (r = .00; p = .95 in sample 3).

    No assumption had been made about the relationship of empathy and educational attainment and corresponding empirical investigations are hard to find. In this re-validation (sample 3), I found a significant correlation of .19.

                The EED short scale has been tested in relation to a number of constructs in order to investigate its validity. Across samples, the EED is highly correlated to its long scale version, the TEQ. Similar to the TEQ, the EED is predominantly a measure of emotional empathy that has a considerable association with (one’s drive for) perspective taking. Furthermore, the EED scale’s associations with authoritarian obedience (i.e. RWA), group-based dominance (i.e. SDO), and prosocial political attitudes confirm the theoretical expectation that individuals with a tendency to draw social ‘borders’ have a lower drive to share others’ emotions. Finally, the EED scale resembles the well established finding that female respondents score higher on self-report measures of empathy than men. All in all, the EED scale demonstrates theoretically coherent associations and can be considered a valid measure of individuals’ drive to share others’ emotions.



    I would like to thank Horst-Alfred Heinrich for his comments on a previous draft of this paper. I also would like to thank the Friedrich-Ebert-Foundation for supporting this research with a grant.


    Wolfgang Karlstetter, Stony Brook University, Department of Political Science, N-723 Social & Behavioral Sciences, 100 Nicolls Rd, NY 11794, USA, E-Mail: