Impairment of affective and cognitive empathy in high functioning autism is mediated by alterations in emotional reactivity

Empathy impairments are considered a key aspect of autism-spectrum disorders (ASD). Previous research consistently shows reduced cognitive empathy, but findings on affective empathy vary, possibly due to experimental design variations (e.g., stimulus modality, social distance) and individual psychological factors (e.g., perceptual abilities, emotional reactivity). This study aims to clarify deficits in affective and cognitive empathy in ASD by addressing these contributing factors. Empathy was examined in 34 autistic individuals and 33 typically developed controls (TDCs) through the Textual Empathy Test (TET). The TET was developed to assess emotional responses when imagining oneself (emotional reactivity) as compared to a target person (friend, stranger) in emotional situations presented via short verbal descriptions. Participants rated emotional states of the target person (cognitive empathy) as well as their own emotional responses when imagining the target person in that situation (affective empathy). Ratings were interpreted relative to normative mean values through standardized regression coefficients. Results showed that high-functioning autism was associated with lower cognitive and affective empathy irrespective of social distance as well as with decreased emotional reactivity compared to controls. Moreover, emotional reactivity mediated the impact of ASD on both empathic components. In summary, altered emotional reactivity may underlie impaired empathy in autistic individuals.


Development and Validation of the Textual Empathy Test
A total of 60 short descriptions of everyday situations including positive (i.e., joy, hope, sexual pleasure, gratefulness, pride, relief; 5 items each) and negative emotions (i.e., shame, anger, disgust, sadness, fear, envy, 5 items each) were adapted from previous studies [1][2][3] or newly added to the existing stimuli set.Subsequently these stimuli were rated by 22 German speaking participants (13 females, 9 males) recruited via convenience sampling using paper pencil questionnaires.The participants' age ranged from 18 to 72 years (m=37.7,sd=±17.8)and the majority was either currently employed or were students.The participants were instructed to imagine themselves in the respective situations.Stimuli were presented in a randomized order.After reading the emotional situation carefully, they had to rate the level of valence of their emotional response on a seven-point emoji-based visual analogue scale from very negative to very positive (A=very positive to G=very negative).Moreover, the participants had to choose one of the 12 emotional categories (joy, hope, sexual arousal, gratefulness, pride, relief, shame, anger, disgust, sadness, fear, envy) most accurately reflecting the emotion evoked by the described everyday situation.Only stimuli which were correctly identified (i.e., emotional category) by at least 70% of participants and had congruent valence ratings qualified for the inclusion to the TET.Not enough items of relief and envy did pass this requirement; therefore, these emotions were not included in the final selection of stimuli.
Table S1 provides an overview of all emotional scenarios in German and its English translation of the close and distant target person conditions.For the self-condition, the pronouns and verb form should be adjusted accordingly.

Stimulus for socially close or distant target persons (marked with respective symbols):
Sie/er geht auf das Konzert ihres/seines Lieblingsmusikers.She/he is going to the concert of her/his favorite musician.

I am going to the concert of my favorite musician.
The self-referential, cognitive and affective ratings for these emotional scenarios were assessed with visual analogue scales from very negative to very positive by the following statements: • Self-referential and cognitive ratings: In der beschriebenen Situation FÜHLT sich DIE PERSON… In the described situation, THIS PERSON feels… TDC-referenced betas.The results of the mixed ANOVA and the Mann-Whitney U test using the control-referenced standardized regression coefficients for empathic responses and emotional reactivity, respectively, reveal similar patterns to the normative standardized regression coefficients reported in the main manuscript (see Figure S1 and compare to Figure 2).Furthermore, autistic individuals had significantly lower empathy-related betas (main effect group: F(1,65)=9.88,p=.003, pη 2 =.13), which were confirmed by Bonferroni-corrected Mann Whitney U tests for the cognitive (U=399, p1tailed = .042,η 2 =.06) as well as the affective component of empathy (U=378, p1tailed = .022,η 2 =.08).There were no significant interactions of group with empathy component and/or social distance (all |F|≤3.68,all p≥.059).Similarly, emotional reactivity betas were significantly lower in the ASD group compared to the TDC group (U=252, p<.001, η 2 =.23).These patterns support the findings of the main analysis using standardized regression coefficient based on an independent sample.Especially, the associations between ASD-related mean self-ratings and individual affective empathic responses as well as emotional reactivity are consistently stronger for TD controls compared to autistic individuals (Bonferroni-corrected Mann-Whitney U tests for cognitive empathy: U=454, p=.36, affective empathy: U=398, p1tailed = .041,η 2 =.06).There were no significant group-related interactions (all |F|≤3.32,all p≥.073).

Unstandardized Regression Coefficients: An Additional Dimension to the Conceptualization of Empathic Responses and Emotional Reactivity
While standardized regression coefficients provide information on the strength/consistency of a linear association between two variables, the unstandardized regression coefficient, the B value, informs about the slope of the relationship.Therefore, complementary analyses (i.e., mixed ANOVA and   ASD-referenced Bs.While the patterns of the ASD-referenced Bs reflect the patterns also observed for the standardized and unstandardized regression coefficient analyses, there was only a significant group difference for emotional reactivity (t(65)=2.92,p = .005,η 2 =.12), with TD controls having significantly higher emotional reactivity compared to autistic individuals (see Figure S5).No such group-related difference reached significance for the empathic response B values (main effect group:

Figure S4. TDC-referenced unstandardized regression coefficients (B) as indicators of empathy, affective empathy and emotional reactivity between TD controls and autistic individuals (ASD). A) shows the cognitive (left) and affective (right) empathy B values representing the slope of the correlation between leave-one-out self-rating means of TD controls and the individual ratings of how another person (socially close in dark blue
F(1,65)=2.42,p=.124).No significant interaction effects with group were detected (all |F|≤0.93,all p≥.339).Coefficients of higher than 1 for the emotional reactivity indicate that when based on ASDrelated standards, TD controls would be classified as emotionally hyperreactive.A similar observation can be made for the cognitive empathy B values, however, here both groups show B vales higher than 1.

Comparability of imagined target persons among autistic individuals and typically developed controls
To inform interpretation of analyses on empathic responses, chi-square and exact Fisher's tests were used to examine the comparability of target persons imagined by the two different groups with respect to type of target person (close: relative/partner, social contact; distant: nobody, stranger from shops/street) and personal distance (close, medium, far).
There was no significant group difference in the type of target persons imagined for the close (Χ 2 (2)=4.46,p=.11) nor for the distant person (p=1.00,Fisher's exact test).For the close person, most participants imagined a social contact with familial relationship (47%), followed by imagining a social contact without familial relationship (27%).The remaining participants (26%) imagined multiple persons including both former groups.For the distant person, the majority thought off a stranger they encountered on the street or in a shop (58%), the remaining participants referred to nobody in particular (41%) or imagined multiple persons from both former groups (1%).
Regarding the level of social distance towards the imagined target person, exact Fisher test did also not reveal a significant difference between the ASD and the TDC group for the close (p=1.00,Fisher's exact test) nor the distant person (p=.43,Fisher's exact test).Across both groups, the relationship for the close person was rated as close (91%) or medium (9%), whereas the relation to the distant person was rated as distant (86%) or medium (11%), respectively.
Figure S1.Comparison of TDC-referenced cognitive empathy, affective empathy and emotional reactivity between TD controls and autistic individuals.A) shows the cognitive (left) and affective (right) empathy beta weights representing the correlation between leave-one-out self-rating means of TD controls and the individual ratings of how another person (socially close in dark blue, socially distant in light blue) would feel in a given emotional situation and how the participants feel when imagining another person in this situation, respectively.The bar plot in B) shows the difference in emotional reactivity (i.e., leave-one-out self-rating means of TD controls to participant's self-rating) between TD controls (red) and autistic individuals (light pink).Error bars indicatethe mean standard error, and jitter represents individual data points per participant.ASD -autism spectrum disorder, TD -typically developed control, *p<.05, ***p<.001.
Figure S2.Comparison of ASD-referenced cognitive empathy, affective empathy and emotional reactivity between TD controls and autistic individuals (ASD).A) shows the cognitive (left) and affective (right) empathy beta weights representing the correlation between leave-one-out self-rating means of autistic individuals and the individual ratings of how another person (socially close in dark blue, socially distant in light blue) would feel in a given emotional situation and how the participants feel when imagining another person in this situation, respectively.The bar plot in B) shows the difference in emotional reactivity (i.e., leave-one-out self-rating means of autistic individuals to participant's self-rating) between TD controls (red) and autistic individuals (light pink).Error bars indicate the mean standard error, and jitter represents individual data points per participant.ASDautism spectrum disorder, TD -typically developed, *p<.05, ***p<.001.
independent t or Mann-Whitney U tests) with the B values could provide additional information on empathic responses and emotional reactivity such as hypo-(B < 1) and hyperresponsivity (B > 1) relative to normative, TDC-referenced and ASD-referenced subjective feelings evoked by the respective emotional situations.Normative Bs.The mixed ANOVA and the independent t-tests with the normative unstandardized regression coefficients (B values) revealed similar patterns for empathic responses and emotional reactivity as previous analyses with the standardized regression coefficients (see FigureS3and compare with Figures2, S1-2).Empathic responses (F(1,65)=3.12,p1tailed=.041,pη 2 =.05) and emotional reactivity (t(65)=3.14,p = .003,η 2 =.13) represented by B values were significantly higher in the TD controls compared to autistic individuals.However, Bonferroni corrected Mann-Whitney U tests did not reveal a statistically significant group difference for cognitive (U=517, p1tailed = .581)and affective empathy (U=427, p1tailed = .093)separately.No significant interactions involving group were found (all |F|≤0.91,all p≥.344).
Figure S4.TDC-referenced unstandardized regression coefficients (B) as indicators of empathy, affective empathy and emotional reactivity between TD controls and autistic individuals (ASD).A) shows the cognitive (left) and affective (right) empathy B values representing the slope of the correlation between leave-one-out self-rating means of TD controls and the individual ratings of how another person (socially close in dark blue, socially distant in light blue) would feel in a given emotional situation and how the participants feel when imagining another person in this situation, respectively.The bar plot in B) shows the difference in emotional reactivity B values (i.e., leave-one-out self-rating means of TD controls to participant's self-rating) between TD controls (red) and autistic individuals (light pink).Error bars indicatethe mean standard error, and jitter represents individual data points per participant.ASD -autism spectrum disorder, TDC -typically developed control, + pmaineffect<.05,**p<.01,***p<.001.
Figure S5.ASD-referenced unstandardized regression coefficients (B) as indicators of cognitive empathy, affective empathy and emotional reactivity between TD controls and autistic individuals (ASD).A) shows the cognitive (left) and affective (right) empathy B values representing the slope of the correlation between leave-one-out self-rating means of autistic individuals and the individual ratings of how another person (socially close in dark blue, socially distant in light blue) would feel in a given emotional situation and how the participants feel when imagining another person in this situation, respectively.The bar plot in B) shows the difference in emotional reactivity B values (i.e., leave-one-out self-rating means of autistic individuals to participant's self-rating) between TD controls (red) and autistic individuals (light pink).Error bars indicate the mean standard error, and jitter represents

Positive Emotions Joy 2. Supplementary Analyses TDC-and ASD-referenced Cognitive Empathy, Affective Empathy and Emotional Reactivity using Standardized Regression Coefficients
empathic responses were then entered into a 2(group: TDC, ASD) by 2(empathy component: cognitive, affective) by 2(personal distance: close, distant) mixed ANOVA model, while the standardized regression coefficients for emotional reactivity were analyzed with a Mann-Whitney U tests due to non-normally distributed data.