Individuals with Autism Share Others' Emotions: Evidence from the Continuous Affective Rating and Empathic Responses (CARER) Task

Abstract

A new task (‘CARER’) was used to test claims of reduced empathy in autistic adults. CARER
measures emotion identification (ability to identify another’s affective state), affective empathy (degree to which another’s affective state causes a matching state in the Empathiser) and affect sharing (degree to which the Empathiser’s state matches the state they attribute to another). After controlling for alexithymia, autistic individuals showed intact affect sharing, emotion identification and affective empathy. Results suggested reduced retrospective socio-emotional processing, likely due to a failure to infer neurotypical mental states. Thus, autism may be associated with difficulties inferring another’s affective state retrospectively, but not with sharing that state. Therefore, when appropriate measures are used, autistic individuals do not show a lack of empathy.

Full text

1
In Press: Journal of Autism and Developmental Disorders
Running Head: EMPATHY IN AUTISM
Individuals with Autism Share Others’ Emotions:
Evidence from the Continuous Affective Rating and Empathic Responses (CARER) Task
Idalmis Santiesteban1, Clare Gibbard2, Hanna Drucks3, Nicola Clayton1, Michael J Banissy4,
Geoffrey Bird5,6
1 Department of Psychology, University of Cambridge, Downing Street, Cambridge, CB2 3EB, UK.
2 University College London Great Ormond Street Institute of Child Health, 30 Guilford Street,
London WC1N 1EH, UK.
3 University of Trier, Universitätsring 15, D-54296 Trier, Germany.
4 Department of Psychology, Goldsmiths, University of London, New Cross, London, SE14 6NW,
UK.
5 Department of Experimental Psychology, University of Oxford, Oxford, OX2 6GG, UK.
6 Social, Genetic & Developmental Psychiatry Centre, Institute of Psychiatry Psychology and
Neuroscience, Kings College London, DeCrespigny Park, London, SE5 8AF, UK.
Correspondence concerning this article should be addressed to Idalmis Santiesteban, Email:
idalmissc@gmail.com

Individuals with Autism Share Others’ Emotions:
Evidence from the Continuous Affective Rating and Empathic Responses (CARER) Task
2
Abstract
A new task (‘CARER’) was used to test claims of reduced empathy in autistic adults. CARER
measures emotion identification (ability to identify another’s affective state), affective empathy
(degree to which another’s affective state causes a matching state in the Empathiser) and affect
sharing (degree to which the Empathiser’s state matches the state they attribute to another). After
controlling for alexithymia, autistic individuals showed intact affect sharing, emotion identification
and affective empathy. Results suggested reduced retrospective socio-emotional processing, likely
due to a failure to infer neurotypical mental states. Thus, autism may be associated with difficulties
inferring another’s affective state retrospectively, but not with sharing that state. Therefore, when
appropriate measures are used, autistic individuals do not show a lack of empathy.
Keywords: autism; empathy; alexithymia; affect sharing; CARER; continuous affective rating;

Individuals with Autism Share Others’ Emotions:
Evidence from the Continuous Affective Rating and Empathic Responses (CARER) Task
3
“People with autism don’t feel empathy”, says a client to a young female hairdresser. The
client tries to disguise her embarrassment when the hairdresser reveals she is actually autistic and
emphatically disagrees with the client’s remark. This real-life conversation, recently witnessed by
one of the authors, reflects the negative, stereotypical view of autism that is not uncommon across
sections of the general population. Its likely origin is in research claiming that an empathy
impairment is a key feature of autism (e.g. Baron-Cohen and Wheelwright, 2004; Decety &
Moriguchi, 2007). Here, it is argued that traditional approaches to the study of empathy could, in part,
be responsible for the negative association between Autism Spectrum Disorder (ASD) and empathic
ability. This study takes an alternative approach in order to examine the reported association between
autism and empathy impairment.
Current Theories of Empathy in Autism
The empathy imbalance hypothesis (Smith, 2009). Smith argues that although autistic
individuals are known to have impairments, such as in theory of mind, that will impact emotion
identification (determining another’s emotional state), they also experience a surfeit of affective
empathy that may be due to increased affect sharing (the process in which attributing an emotional
state to another causes you to share that state). Smith says that as a consequence, those with ASD are
susceptible to empathic overarousal, which in turns lead to personal distress. When overwhelmed by
the affective state of another, autistic individuals become unable to produce appropriate empathic
behaviours (i.e. behaviours which successfully ameliorate the negative state of the empathic target,
henceforth, ‘Target’), which Smith argues has given rise to the (incorrect) view that individuals with
autism lack empathy. However, this interesting theory has thus far received mixed empirical support;
while some findings support intact affective empathy in ASD, accompanied by impairment in
emotion identification (e.g. Dziobek et al., 2008; Rueda, Fernández-Barrocal & Baron-Cohen, 2014),
others do not support Smith’s claims (e.g. Shamay-Tsoory, Tomer, Yaniv & Aharon-Peretz, 2002;
Lawrence et al., 2004; Baron-Cohen et al., 2003; Adler, Dvash & Shamay‐Tsoory, 2015). Overall,

Individuals with Autism Share Others’ Emotions:
Evidence from the Continuous Affective Rating and Empathic Responses (CARER) Task
4
Smith’s theory could benefit from direct empirical testing. The current study, which employs a new
empathy task – see description below – could provide an important test of this theory.
The role of alexithymia in autism and empathy (Bird et al., 2010; Bird & Cook, 2013). The
second account relevant to empathic ability in ASD relates to the high comorbidity of alexithymia and
autism. The term Alexithymia, literally meaning ‘without words for emotions’, is used to describe a
cluster of cognitive traits including an inability to identify and describe one’s own feelings, and
externally-oriented thinking (Sifneos, 1973). The estimated prevalence of alexithymia among the
general population is 10% (Salminen et al., 1999). However, previous studies of alexithymia in the
autistic population suggest an increased prevalence, with approximately 50% of ASD adults reporting
severe alexithymia (Hill, Berthoz, Frith, 2004; Bird & Cook, 2013). Furthermore, recent experiments
investigating the extent to which alexithymia can predict some of the socio-emotional impairments
commonly associated with ASD have shown that: i) reduced eye-fixation (Bird, Press & Richardson,
2011), ii) poor recognition of emotional facial and vocal expressions (Cook, Brewer, Shah & Bird,
2013, Heaton et al., 2012; Trevisan et al., 2016; Bothe, Palermo, Rhodes, Burton, & Jeffery, 2019), and,
iii) reduced empathic activation of the insula (Bird et al, 2010; Silani et al., 2008) is explained by
alexithymia, and that no autism impairment in these processes is observed after controlling for
alexithymia. Thus, the ‘alexithymia hypothesis’ (Bird & Cook, 2013) suggests that the ‘emotional
symptoms of autism’ (including the claim of reduced empathy in autism) are a product of co-occurring
alexithymia. The hypothesis suggests that sampling variance with respect to alexithymia in studies of
autistic individuals can explain both the inconsistencies in the results of experimental investigations of
emotional processes in autism, and the heterogeneity across individuals with autism with respect to
emotional ability. Thus, the alexithymia hypothesis would suggest that apparent empathy impairments
in autism should be reduced, or even negated entirely, when alexithymia is controlled for.
Bird & Viding’s (2014) Self to Other Model of Empathy (SOME) expands on this hypothesis
with respect to the impact of alexithymia and autism on empathy. They argue that autism is associated
with theory of mind impairments (e.g. Baron-Cohen, 2000; Frith, 1994, 2012) but not with difficulties

Individuals with Autism Share Others’ Emotions:
Evidence from the Continuous Affective Rating and Empathic Responses (CARER) Task
5
recognising facial or vocal expressions of emotion. Thus, one would expect differential contributions
of autism and alexithymia to impairments in emotion identification depending on the degree to which
theory of mind (autism) or facial/vocal emotion recognition (alexithymia) is required to determine the
Target’s state. Once identified, it is argued that the degree to which the Target’s state affects the
Empathiser’s state (affect sharing) will solely be affected by alexithymia, not autism (Bird & Viding,
2014; Bird & Cook, 2013; Bird et al., 2010). Accordingly, levels of alexithymia were recorded in the
current experiment in order to see whether any empathy-related impairments in autism can be
accounted for by alexithymia.
Current Empathy Measures
Although our ability to empathise with others is thought to lie at the heart of successful social
interaction, researchers are yet to agree upon what empathy is (e.g. Batson, 2009), or how to measure
it. The existing definitions of empathy range from arguably simpler processes such as recognition of
emotional facial expressions and emotional contagion, to more complex forms requiring the
Empathiser to recognise that their affective state is caused by the emotional state of the empathic
target (de Vignemont & Singer, 2006). The lack of an agreed-upon definition is accompanied by a
lack of consistent experimental methodology. The most frequently used measures of empathy rely on
participants’ self-report, with two of the most common being the Interpersonal Reactivity Index (IRI,
Davis, 1980) and the Empathy Quotient (EQ, Baron-Cohen & Wheelwright, 2004). However, relying
on self-report is particularly problematic to assess the claim of empathy deficits in autism: individuals
with autism are continuously told they lack empathy, and so to report typical empathy implies that
they think the collection of medical professionals and scientific researchers claiming empathy deficits
in autism are wrong, or that their diagnosis of autism (which may be an intrinsic part of their self-
identity and/or linked to financial or other support) is wrong.
Recognition that the use of self-report measures can sometimes be problematic has prompted
researchers to devise various types of behavioural or neurophysiological tasks, using a range of
stimuli, that purport to allow empathy to be measured. For example, some studies use images of

Individuals with Autism Share Others’ Emotions:
Evidence from the Continuous Affective Rating and Empathic Responses (CARER) Task
6
facial expressions in which participants (Empathisers) are asked to identify the Target’s emotion and
to report their own level of arousal and concern after viewing each image (e.g. Dziobek et al, 2008).
While the former can be considered a measure of ‘emotion identification’ (the ability to identify the
affective state of others), the latter is interpreted as a measure of emotional/ affective empathy (the
extent to which the Target’s affective state causes the Empathiser’s state to match the Target’s
emotion, Coll et al., 2017). Other studies, particularly those measuring empathy for pain, record
participants’ motor-evoked potentials (MEPs) following Transcranial Magnetic Stimulation while
they observe images of body parts (hands / feet) in painful or non-painful situations (e.g. Avenanti et
al., 2005; Minio-Paluello et al, 2009). A reduction in the amplitude of the motor-evoked potential
recorded from the Empathiser’s hand is interpreted as an empathic response.
What about Individual Differences in Empathy?
As can be seen, previous research has focused on measuring empathy as the outcome of a
process: empathy has occurred if the Empathiser is in the same state as the Target as a consequence
of the Target’s state. Coll and colleagues (2017) have argued that such a conceptualisation results in a
binary outcome – empathy either occurs or it does not – and the reasons for any ‘empathy failure’ are
unclear. Instead, they argued that empathy should be viewed as a process with (at least) two
constituent parts: emotion identification and affect sharing. Individual differences in the former
process reflect the degree to which one can accurately infer the other’s emotional state, while
individual differences in the latter reflect individual differences in the degree to which attribution of
an affective state to the other causes the same state to be instantiated in the self. Under this definition,
a person’s empathic abilities reflect the extent to which they feel an emotion that matches the
emotion they identified in the Target, even when that emotion differs from what the Target actually
feels. This is a significant departure from the traditional view of empathy in which the actual emotion
of the Target is the reference point from which the Empathiser’s state is judged. Following their
definition, Coll et al. suggest that one appropriate measure of affect sharing is the degree to which the
state elicited in the Empathiser matches the state the Empathiser identifies in the Target. For example,

Individuals with Autism Share Others’ Emotions:
Evidence from the Continuous Affective Rating and Empathic Responses (CARER) Task
7
if Empathiser A has an affective response closer to what A believes to be the affective state of the
Target than Empathiser B has to what B believes to be the affective state of the Target, then
Empathiser A could be considered more empathic (more formally, they would be considered to have
a greater degree of affect sharing).
The perspective offered by Coll and colleagues is a radical departure from both prior
empathy research and the lay understanding of empathy. As an extreme example, an individual could
be considered empathic (or at least to have typical affect sharing) even when they show the opposite
emotion to what the Target is feeling, if their state matches the emotion they identified in the Target.
However, adopting Coll and colleagues’ methodological recommendation for the analysis of
empathic ability enables measurement of individual (or group) differences in the processes necessary
to produce an empathic response. For completeness therefore, this study will measure the traditional
concept of ‘affective empathy’ (which can be defined as the degree to which the Target’s state causes
the Empathiser to be in matching state) in addition to emotion identification (the accuracy with which
an Empathiser can determine the Target’s state; sometimes called cognitive empathy), and affect
sharing (the degree to which the Empathiser’s state matches the state they identify in the Target). All
three measures will be used to compare neurotypical individuals and individuals with autism using a
new empathy task as described below.
The Continuous Affective Rating and Empathic Response (CARER) Task
The CARER task is an extension of the Empathic Accuracy Task (EAT, e.g. Ickes, Stinson,
Bissonnette & Garcia, 1990; Zaki, Bolger & Ochsner, 2008). The EAT consists of two phases. In the
first phase, individual Targets were invited to the lab to record short video clips describing real-life
emotional events. The targets then rated, on a continuous scale, how they felt while telling each story.
In the second phase, participants watched these videos and provided continuous ratings of how they
thought the target was feeling. The key measure of the EAT task was empathic accuracy, calculated
by correlating the Empathisers’ ratings of how they thought the Targets felt and the Targets’ ratings
of their own feelings.

Individuals with Autism Share Others’ Emotions:
Evidence from the Continuous Affective Rating and Empathic Responses (CARER) Task
8
The new (CARER) task – See Figure 1 – expands on the EAT by including additional
variables that enable a better understanding of the various aspects of empathy. First, CARER includes
emotionally neutral stories to enable comparisons with the affective stories. Second, Targets in the
CARER task were adults representative of the local community in terms of age, gender and ethnicity.
Third, in addition to ratings of how they thought the Target was feeling (‘Other’ ratings), Empathisers
were also asked to rate their own feelings in response to each of the stories (‘Self’ ratings). Fourth, in
most empathy studies, affective ratings are collected from Empathisers following exposure to
emotional stimuli (offline) and consist of single ratings. Therefore, in order to compare our data with
previous studies, we also asked Empathisers to provide single/offline ratings after watching each
video. In sum, the CARER task included two blocks. In the first block, Empathisers provided
continuous ratings of their own feelings (‘Online Self’) while watching the video, then once the video
had finished, they rated how they thought the Target was feeling (‘Offline Other’). In the second
block, the order of ratings was reversed (first ‘Online Other’ followed by ‘Offline Self’). The order of
blocks was randomised across participants, who completed both blocks in succession. This task
design allows emotion identification, affective empathy as classically defined, and affect sharing to
be measured. Emotion identification is assessed by comparing the Empathisers’ ratings of the
Targets’ affective state to the ratings provided by the Targets themselves. Affective empathy is
assessed by comparing the extent to which Empathisers’ self-ratings correspond to the Targets’ own
ratings. Affect sharing is assessed using the correspondence between the Empathiser’s self-ratings
and their rating of the Target’s state.
The use of both on- and off-line measures is an important feature of the CARER task. The
online measure resembles the dynamics involved in real life social interactions, in which the
Empathiser is required to respond to the Target’s state instinctively, at speeds approaching real time.
In contrast, the offline measures (as used in most empathy tasks) are a product of more reflective
processing in which the Empathiser is more of an observer. Such offline reflection has been argued to
rely on the kind of explicit, verbally-mediated theory of mind which has been shown to be difficult

Individuals with Autism Share Others’ Emotions:
Evidence from the Continuous Affective Rating and Empathic Responses (CARER) Task
9
for individuals with autism (Baron-Cohen, Leslie & Frith, 1985 - see Schilbach, 2014; Schilbach et
al., 2013 for an in-depth discussion of online vs offline social cognition).
In summary, this study adopts a new methodological approach to the study of empathy in
autism. Using a novel empathy task, affective empathy as classically defined (the degree to which the
Empathiser’s affective state matches that of the Target), emotion identification (the degree to which
the Empathiser can accurately identify the Target’s affective state), and affect sharing (the degree to
which the Empathiser’s affective state matches that they identify in the Target), can all be measured.
To the degree that alexithymia can explain any deficits in these measures in autism, one would expect
any apparent autism-related deficits to be reduced to non-significance after controlling for
alexithymia.
Method
Participants
Due to the difficulties inherent in recruiting individuals with a condition with low prevalence
such as autism, we used an opportunity sampling method. To avoid sole reliance on null hypothesis
significance testing, we (a) perform additional Bayesian analysis; and (b) report effect size measures
in the Results section. Our autism group consisted of 21 individuals with an independent clinical
diagnosis of autism (13 female, age range: 18-55, Mean age =29, SD = 2.06) and 45 adults without an
autism diagnosis (29 female, age range: 18 -53, Mean age =25, SD = 1.22) who all volunteered to
take part in the study in exchange for a small monetary reimbursement. The groups did not differ in
terms of age, t(34.5) = 1.82, p = .08, or gender χ2 (1) = .04, p = .84. In light of the memory component
of the empathy task – recalling a story during the offline rating condition – we used the logical
memory subscale of the Weschler’s scale, fourth edition (WMS-IV, Wechsler, 2009), to measure
immediate recall. No group differences were found on this measure [t(63) = -.17, p = .86; ASD:
M=26.38, SD = 7.24; Controls: M = 26.68, SD = 5.96].
Due to the nature of the empathy task, which involved exposure to emotionally charged stories,
participants were required to complete the Beck’s Depression Inventory (BDI-II; Beck, Steer &

Individuals with Autism Share Others’ Emotions:
Evidence from the Continuous Affective Rating and Empathic Responses (CARER) Task
10
Brown, 1996) prior to taking part in the experiment. Only those scoring below 30 (the cut off for
clinical depression) on the BDI were invited to take part in the experiment. Since the BDI measures
depressive symptoms over a two-week timeframe, it was administered again at the beginning of the
experimental session in the lab. No participants were excluded on the basis of BDI, however, there
was a significant group difference on BDI scores, t(29.89) = 2.74, p = .01, d = .76; ASD M =10.48,
Range = 0 – 24; SD = 7.2; Controls: M = 5.68, Range = 0 – 20, SD = 5.15. Ethical approval was
granted by the Cambridge Psychological Research Ethics Committee, and all participants provided
written informed consent.
[Insert Table 1 about here]
Table 1 – Demographics of ASD Group
Participant
Gender
Age
AQ Score
BDI
Logical
Memory
TAS-20
1
M
22
36
7
32
50
2
F
20
36
5
27
46
3
M
30
42
3
22
63
4
M
24
49
20
39
74
5
F
40
30
18
31
65
6
F
18
42
17
38
72
7
F
29
47
15
31
68
8
M
33
48
24
19
80
9
F
23
45
11
20
65
10
F
21
16
21
18
72
11
M
40
8
14
22
69
12
F
19
38
5
28
61
13
F
27
11
0
38
61
14
F
24
45
12
21
76
15
F
23
42
15
29
78
16
F
22
33
10
27
46
17
M
41
39
3
17
61
18
F
55
39
2
28
59
19
M
33
26
0
14
60
20
F
39
32
12
21
63
21
M
26
30
6
32
70
Group Means
29
35
10.48
26.38
64.71

Individuals with Autism Share Others’ Emotions:
Evidence from the Continuous Affective Rating and Empathic Responses (CARER) Task
11
Materials and Procedures
Empathy task (CARER). As described above, this task consisted of two phases. In phase 1, we
invited adult volunteers (Targets) to the lab. Each Target recorded two brief videos (approximately 30
seconds each). Prior to the recordings, Targets were instructed that one of the videos should describe a
real-life event when they felt particularly sad (affective story) and the second video should also be a
real-life event but devoid of affective language, for example, a description of their journey to work, a
daily routine, what they do for a living, what subject they are studying, etc. (neutral story). Targets
had full control over the events they chose to describe, that is, the stories were not scripted. Following
the video recordings, Targets were asked to watch each video and provide continuous ratings of how
they felt when they were telling the story. The continuous rating scale was presented vertically next to
the video display and the values ranged from 0 (extremely calm) to 10 (extremely upset).
Figure 1 shows the sequence of the second phase of the CARER task. During this phase,
participants (Empathisers) were presented with the videos described above. A total of 32 trials (16
affective and 16 neutral) were presented in two blocks of 16 trials each in a pseudorandomized order,
such that no more than three trials of the same type were presented in succession. In Continuous Self
blocks participants were asked to rate how they feel while watching each video (Online Self rating)
and once the video finished, they were asked to rate how the Target felt (Offline Other rating), using a
scale from 0 (extremely calm) to 10 (extremely upset). During Online Other blocks the rating was for
the Target (how is the person in the video feeling?) and the offline rating was for the self (how do you
feel?). The online ratings were continuous, participants could provide as many ratings as they wished,
so that their real-time response to the Target could be recorded. The order in which the blocks were
completed was counterbalanced across participants. After each video, participants were asked to
record a brief video message to the Target based on the story they just watched. The video responses
were scored independently by researchers blind to the experimental design and participant group. The
researchers coded the following variables: frequency of empathic phrases (e.g. I’m sorry to hear…,
that sounds awful, I feel for you, I can’t image…, etc), frequency of verbal signs of personal distress
(e.g. soft tone of voice, breaking voice, slow speech, sadness), and frequency of non-verbal signs of

Individuals with Autism Share Others’ Emotions:
Evidence from the Continuous Affective Rating and Empathic Responses (CARER) Task
12
distress (e.g. teary, sad facial expressions – turned down mouth eyebrows raised/knitted –, hand over
mouth or chest). The task took approximately 35-40 minutes to complete.
[Insert Figure 1 about `]
Figure 1. Example of a trial from the Continuous Self block of the Continuous Affective Ratings and
Empathic Response (CARER) task. A control trial would contain a neutral story devoid of emotional
language, for example, a description of a routine journey into work. Continuous Other blocks required
online continuous ratings of how the Target feels, and an offline rating of how the Empathiser
(participant) feels.
Self-report measures
Following the empathy task, participants completed 3 self-report questionnaires: the
Interpersonal Reactivity Index (IRI; Davis, 1980) which consists of 4 subscales: empathic concern,
fantasy, personal distress, and perspective taking to measure empathy; the Autism Quotient (AQ;
Baron-Cohen et al., 2001) to measure autistic traits; and the Toronto Alexithymia Scale (TAS-20;
Bagby, Taylor & Parker, 1994) to measure alexithymia. The TAS-20 consists of 3 subscales,
difficulty describing feelings, difficulty identifying feelings and externally-oriented thinking. The
entire experimental session lasted approximately 80-90 minutes.

Individuals with Autism Share Others’ Emotions:
Evidence from the Continuous Affective Rating and Empathic Responses (CARER) Task
13
Data analysis
The CARER task enables three dependent variables to be calculated: (i) Emotion
Identification (Target’s self rating– Empathiser rating of the Target’s state) where lower scores
indicate more accurate Emotion Identification; (ii) Affective Empathy (Target’s self rating –
Empathiser’s self rating) – where lower scores indicate higher Affective Empathy
1
; and (iii) Affect
Sharing (Empathiser’s rating of the Target’s state – Empathiser’s self rating) – where lower scores
indicate higher Affect Sharing. These three values can be calculated from both online and offline
ratings.
With respect to online ratings, in principle participants’ ratings could be treated as a
continuous time series. However, a preliminary inspection of the data revealed that the rating data
showed a consistent pattern whereby ratings were low in the first half of videos (where Targets were
providing the background to the emotional event) and higher in the latter stages of the video (where
the emotional climax of the video occurred). Thus, online rating data were split into two epochs (first
and last 15 seconds of each video) and averaged within epoch. For offline analyses, participants
provided a single score while Target ratings were continuous scores. Debriefing confirmed that all
participants provided the highest affective rating for each story as their offline rating, therefore, we
subtracted the participants’ offline ratings from the Target’s maximum rating. Thus, nine dependent
variables were obtained: Early and Late Online, plus offline versions of Emotion Identification,
Affective Empathy and Affect Sharing scores.
The data were analysed with SPSS (version 26). The Online data were analysed using
ANCOVA with Story type (affective vs control) and Time (early vs late) as within-subject factors,
Group (autism vs control) as a between-subjects factor and Alexithymia scores as a covariate. The
analysis of the offline data included the same factors, with the exception of Time. Where sphericity
1
Negative scores would indicate ‘excessive’ Affective Empathy, showing that the Empathiser’s reaction to the
Target’s affective state is stronger than the Target’s own reaction.

Individuals with Autism Share Others’ Emotions:
Evidence from the Continuous Affective Rating and Empathic Responses (CARER) Task
14
assumptions were not met, Greenhouse-Geisser corrected values are reported. Bonferroni corrections
were used for post-hoc multiple comparisons.
A series of subsequent multiple regression analyses identified the unique variance explained
in each CARER variable by autism, alexithymia and gender. Accordingly, multiple regression models
included TAS scores, AQ scores and gender (note that the pattern of significance was the same if
diagnostic status was coded as a categorical binary variable).
Results
Individual Differences Approach Accounting for Alexithymia
Emotion Identification Online. This analysis revealed no main effects or interaction reached
significance
2
, all ps ≥ .083. We further tested the strength of evidence for the null hypothesis of no
interaction between Story type and Group with a Bayesian analysis using JASP (JASP Team, 2019).
A Bayes Factor (BF10) above 3 indicates substantial support for H1 and a BF less than 0.33 indicates
substantial support for H0 (Jeffrey, 1961; Raftery, 1995). JASP default priors were used as the model
for H1. Bayes Factors of 1.27 and 0.28 were obtained for the two- (Story type x Group) and three-
way (Story type x Group x Time) interactions respectively. Thus, Bayesian analyses provided little
support for either the null or alternative hypotheses with respect to the two-way interaction, and
substantial support for the null with respect to the three-way interaction.
Emotion Identification Offline. Figure 2 (Panel A) shows the offline data for Emotion
Identification. We found no main effects of Story type (p = .36), or Group (p = .081). However, the
2
When checking for homogeneity of variance using Levene’s test, the variable affective emotion
identification was significant at T2. Therefore, findings from the ANCOVA relating to the crucial Group x
Story type interaction were re-assessed using the non-parametric Kruskal-Wallis test in JASP (JASP Team,
2019). We calculated a difference score between affective and neutral trials to obtain a single difference
score and compared this score between the groups. The results of this analysis showed the same pattern of
significance as the ANCOVA results (X2(1) =.2, p = .655).

Individuals with Autism Share Others’ Emotions:
Evidence from the Continuous Affective Rating and Empathic Responses (CARER) Task
15
Story type × Group interaction was significant
3
, F(1,63) = 5.64, p = .021, η2p = .08, 90% CI [.006,
.201]. Post-hoc tests revealed that this was driven by a reduction in emotion identification in the
autistic group on affective trials (M = 2.69, SEM = .53), compared to controls (M = 1.11, SEM = .31;
F(1,63) = 5.15, p = .027, η2p = .08, 90% CI [.005, .192]. The Group contrast on neutral trials was not
significant, p = .92. A Bayes Factor of 6.915 was obtained for the two-way (Story type x Group)
interaction, providing substantial support for the alternative hypothesis.
Affective Empathy Online. No main effects or interactions reached significance, all ps ≥ .10.
Bayes Factors of 1.426 and 2.13 were obtained for the two- (Story type x Group) and three-way
(Story type x Group x Time) interactions. Thus, Bayesian analyses provided little support for either
the null or alternative hypotheses with respect to both the two- and three-way interactions.
Affective Empathy Offline. Figure 2 (Panel B) shows the offline data. The analysis revealed a
significant Group × Story type interaction, F(1,63) = 3.98, p =.05, η2p =.06, 90% CI [.000, .171].
Pairwise comparisons showed that while there were no Group differences on neutral stories (p = .43),
significant differences between the Groups were present on affective stories, with reduced affective
empathy scores in the autism Group (M = 4.58, SEM = .68) compared to the control Group (M = 2.60,
SEM = .40); F(1,63) = 4.79, p =.032, η2p =.07, 90% CI [.003, .186]. A Bayes Factor of 6.59 was
obtained for the two-way (Story type x Group) interaction, providing substantial support for the
alternative hypothesis.
Affect Sharing Online. This analysis revealed no significant main effects or interactions, all
ps ≥ .12. Bayes Factors of 0.38 and 0.39 were obtained for the two- (Story type x Group) and three-
way (Story type x Group x Time) interactions respectively. Thus, Bayesian analyses provided
anecdotal support for the null hypothesis with respect to both the two- and three-way interactions.
3
The Levene’s test for the variable emotion identification for affective stories was significant, therefore, the
Group x Story type interaction found in the ANCOVA was further assessed with a non-parametric Kruskal-
Wallis test. We calculated a difference scores between affective and neutral stories to obtain a single difference
score and compared this score between the groups. This analysis supported the ANCOVA results, (X2(1) = 6.88,
p = .009).

Individuals with Autism Share Others’ Emotions:
Evidence from the Continuous Affective Rating and Empathic Responses (CARER) Task
16
Affect Sharing Offline. Figure 2 (Panel C) shows the offline data for Affect Sharing. In
common with the online analysis, the offline data showed no significant main effects or interactions,
Crucially, in contrast to the offline Emotion Identification and Affective Empathy analyses, the
Group × Story type interaction failed to reach significance (p = .85). Again, we carried out a
Bayesian analysis, which yielded a BF of 0.593, thus providing little support for either the null or
alternative hypothesis.
[Insert Fig. 2 about here]
Figure 2. Offline data for Emotion Identification (Panel A, calculated by subtracting
participants’ ratings of how the Targets feels from the Targets’ own ratings), Affective Empathy
(Panel B, calculated as the difference between participants’ self ratings and the Targets’ own ratings)
and Affect Sharing (Panel C, calculated by subtracting the Empathiser’s own state (self ratings) from
Empathiser’s judgement of the Target’s state (other ratings). In all measures, lower scores indicate
better performance of the measure in question. Error bars represent the SEM.
Video Responses. Figure 3 shows the box plots for each of the measures derived from the
video responses: frequency of empathic phrases (e.g. I’m sorry to hear…, that sounds awful, I feel for
you, I can’t image…, etc), frequency of verbal signs of personal distress (e.g. soft, breaking, slow,

Individuals with Autism Share Others’ Emotions:
Evidence from the Continuous Affective Rating and Empathic Responses (CARER) Task
17
sadness), and frequency of physical signs of distress (e.g. teary, sad facial expressions – turned down
mouth eyebrows raised/knitted –, hand over mouth or chest). Two researchers blind to the
experimental design independently scored the participants’ video responses. One researcher coded all
video responses from the whole sample, to ensure reliability of these ratings a second researcher
scored the responses of 46 randomly selected participants. Both coders were mostly unaware if
participants were autistic or not, however, two participants explicitly mentioned in the videos that
they had an ASD diagnosis. We tested inter-rater reliability with intraclass correlation coefficients
(ICC) estimates and their 95% confidence intervals, which were calculated based on a median rating
(k=2), absolute agreement, 2-way mixed effects model. A high degree of ICC reliability was found
for the three measures (frequency of empathic phrases, ICC = .95, 95% CI = .91 to .97, F(45,45) =
21.30, p <.001); frequency of verbal signs of personal distress, ICC = .97, 95% CI = .94 to .98,
F(45,45) = 35.07, p <.001); frequency of physical signs of distress ICC = .85, 95% CI = .73 to .92,
F(45,45) = 6.63, p <.001).
Mann-Whitney tests on the video response variables indicated group differences during
affective trials for frequency of empathic phrases used, U = 283, p = .005, the ASD group (Mean
Ranks = 24.48) used fewer empathic phrases than controls (Mean Ranks = 37.07). Group differences
were also found in the frequency of verbal signs of personal distress U = 219, p < .001 (Mean Ranks
ASD = 21.45, controls = 38.51) and frequency of physical signs of distress U = 299, p = .005 (Mean
Ranks ASD = 25.24, controls = 36.70). Notably, the direction of the group differences in the
measures of personal distress, both for verbal and non-verbal cues – see Fig. 6 – do not support the
empathy imbalance hypothesis’ prediction that ASD individuals’ potential empathy overload leads to
higher levels of personal distress. Our data show the opposite pattern, with ASD individuals
displaying fewer signs of personal distress than neurotypicals. No significant group differences were
found on any these measures for the neutral videos, all ps >.05.
Since significant differences between the groups were found in the video responses, the
correlations between these measures, AQ and alexithymia scores were further explored. For

Individuals with Autism Share Others’ Emotions:
Evidence from the Continuous Affective Rating and Empathic Responses (CARER) Task
18
alexithymia scores, the TAS subscales (difficulty identifying feelings, difficulty describing feelings
and externally-oriented thinking) were deemed to be more informative in this analysis than the total
alexithymia score. AQ scores correlated with all three measures: frequency of empathic phrases [r
(63) = -.27, p = .03]; frequency of verbal signs of personal distress [r (63) = -.26, p = .041] and
frequency of physical signs of distress [r (63) = -.29, p = .023]. Furthermore, significant correlations
between video responses and the alexithymia subscales were also present. For example, difficulty
describing feelings was correlated with frequency of empathic phrases [r (64) = -.28, p = .027];
frequency of verbal signs of personal distress [r (64) = -.27, p = .031] and frequency of physical signs
of distress [r (64) = -.34, p = .005]. For difficulty identifying feelings the only significant correlation
was with frequency of physical signs of distress [r (64) = -.35, p = .005]; whereas no significant
correlations were found between externally-oriented thinking and any of the video responses.
[Insert Figure 3 about here]
Figure 3. Box plots of each of the measures used in the analysis of video responses: A)
Number of empathic phrases, B) Verbal signs of distress, C) Physical signs of distress. The bottom
and top whiskers represent the minimum (Q1) and maximum (Q3) values respectively. The diagram
also shows the median (asterisk) and mean (circle) observation for each group.
Self-Report Measures. Due to technical difficulties one participant’s AQ score, and a second
participant’s IRI score were not recorded, both participants were from the control group. As expected,
the groups differed in their AQ score, t(63) = 7.16, p <.001, d = 1.89, 95% CI [1.28, 2.50], (ASD M =
35, SEM = 2.54; controls M = 16, SEM = 1.34). The Alexithymia scores were also significantly

Individuals with Autism Share Others’ Emotions:
Evidence from the Continuous Affective Rating and Empathic Responses (CARER) Task
19
different between the groups, t(64) = 8.83, p <.001, d = 2.33, 95% CI [1.67, 2.98]; with higher scores
found for the ASD group (M = 64.71, SEM = 2.07) compared to controls (M = 43.33, SEM = 1.34).
This pattern was repeated for three of the four IRI subscales, empathic concern, t(63) = -4.74, p
<.001, d = 1.25, 95% CI [.69, 1.81], (autistics: M = 14.67, SEM = 1.29; controls: M = 20.84, SEM
=.66); perspective taking, t(63) = -4.13, p <.001, d = 1.09, 95% CI [.54, 1.64], (autistics: M = 13.33,
SEM = 1.03; controls: M = 18.86, SEM =.78); and fantasy, t(63) = -4.12, p <.001, d = 1.08, 95% CI
[.53, 1.64], (autistics: M = 11.62, SEM = 1.33; controls: M = 17.43, SEM =.74). The personal distress
subscale scores did not significantly differ between the groups (autistics: M = 14, SEM = 1.43;
controls: M = 11.7, SEM =.70, p = .16).
Correlations. Table 1 shows the correlations between all dependent variables across all
participants. Of particular interest were correlations with the key variables from the CARER task:
affective empathy, emotion identification and affect sharing, both online and offline. With respect to
the study hypotheses, we found that autism traits (AQ) were negatively correlated with emotion
identification offline, r(65) =.40, p = 001, and with affective empathy, both online - r(65) = .31, p =
.01 – and offline - r(65) =.43, p <.001. Furthermore, AQ scores were positively correlated with
alexithymia scores, r(65) = .66, p <.001 (and all TAS subscales) and negatively correlated with the
following IRI subscales: empathic concern, r(64) = -.40, p = .001; perspective taking, r(64) = -.34, p
= .006; and fantasy, r(64) = -.403, p = .001. However, contrary to the prediction from the empathy
imbalance hypothesis in autism, we found no correlation between AQ scores and the IRI personal
distress subscale, r(64) = .097, p = .447.
Notably, we also found a negative correlation between overall alexithymia scores (TAS) and
affective empathy both online – r(66) = .28, p = .025 – and offline - r(66) =.34, p = .006; and with
affect sharing offline, r(66) = .28, p = .022. Total alexithymia scores were also negatively correlated

Individuals with Autism Share Others’ Emotions:
Evidence from the Continuous Affective Rating and Empathic Responses (CARER) Task
20
with all the IRI subscales: empathic concern r(65) = -.62, p < .001; perspective taking, r(65) = -.48, p
< .001, fantasy (r(65) = -.478, p < .001) and personal distress (r(65) = .266, p = .032.
Multiple Regressions. Finally, a multiple regression analysis was performed on each of the
CARER variables which included the following predictors: AQ scores, TAS scores and Gender in an
attempt to understand the contribution of each of these independent variables on the performance of
the CARER task (see Supplemental Materials for full model information). Tests for
multicollinearity indicated that a very low level of multicollinearity was present (VIF = 1.80 for AQ
scores, 1.79 for TAS scores, and 1.04 for Gender). None of the models, or independent variables
within the models, were significant predictors of the online measures. However, the model was a
significant predictor of offline emotion identification, F(3, 64) = 4.28, p = .008, R2 = .174, R2 adjusted =
.133, and out of the three independent variables AQ was the only significant predictor, (b = .07,
t(.02) = 3.19, p = .002). The model predicting offline affective empathy was also significant, F(3,
64) = 6.29, p = .001, R2 = .236, R2 adjusted =.199. In this model, AQ was a significant predictor, (b =
.08, t(.08) = 2.69, p = .009). Finally, the model for predicting affect sharing offline was also
significant, F(3, 64) = 4.20, p = .009, R2 = .17, R2 adjusted =.13. In this model, Gender was a significant
predictor, (b = 1.18, t(.44) = 2.68, p = .009) and alexithymia scores were a marginally significant
predictor, (b = .04, t(.02) = 1.88, p = .065).

21
21
Table 2. Correlations between key variables from the CARER task and self-reported questionnaires
1
2
3
4
5
6
7
8
9
10
11
12
13.
14
15
1. Emotion Ident Online
2. Emotion Ident. Offline
.26*
3. Affect. Empathy
Online
.28*
.40**
4. Affect. Empathy
Offline
.08
.70**
.48**
5. Affect Sharing Online
-.26*
-.16
.33**
-.02
6. Affect Sharing Offline
-.16
-.06
.26*
.67**
.14
7. AQ
.09
.40**
.31*
.43**
-.13
.18
8. TAS-20
.15
.18
.28*
.34**
-.13
.28*
.66**
9. TAS – DDF
.08
.16
.25*
.37**
-.02
.35**
.65**
.90**
10. TAS – DIF
.10
.17
.20
.29*
-12
.23
.63**
.88**
.72**
11. TAS – EOT
.21
.11
.24*
.13
-.20
.08
.28*
.63**
.44**
.27*
12. IRI Fantasy
-.16
-.11
-.30*
-.28*
-.04
-.28*
-.40**
-.48**
.48**
-.27*
-.48**
13. IRI Empathic Concern
-.14
-.40**
-.39**
-.37**
.02
-.12
-.40**
-.62**
-.58**
-.48**
-.48**
.50**
14. IRI Perspective
Taking
-.03
-.32**
-.40**
.36**
.02
-.17
-.34**
-.48**
-.36**
-.45**
-.37**
.35**
.67**
15. IRI Personal Distress
.16
-.22
-.13
-.24
-.15
-.12
.01
.27*
.19
.34**
.07
.14
.01
.02
** Correlation is significant at the 0.01 level (2-tailed).
* Correlation is significant at the 0.05 level (2-tailed).
Abbreviations: AQ = Autism Quotient, TAS-20 = Toronto Alexithymia Scale - 20 items, DDF = difficulty describing feelings, DIF = difficulty identifying
feelings, EOT = externally oriented thinking, IRI = Interpersonal Reactivity Index.

22
22
Discussion
The findings of this study suggest that adopting a new methodological approach to empathy
research can challenge previous claims that individuals with autism lack empathy. Past research has
focused on measuring affective empathy - the degree to which the Empathiser’s state matches that of
a Target – which is the outcome of a number of cognitive processes. It has been suggested that it may
be more useful to examine those constituent processes which may, or may not, result in an outcome
which would be defined as affective empathy, in order to examine the empathic abilities of a group of
individuals (Coll et al., 2017). This was the approach taken in this study; a group of individuals with
autism and a control group completed a novel task (CARER) which provided a measure of classic
affective empathy, and also emotion identification – the ability to identify accurately another’s
affective state –, and affect sharing – the degree to which the Empathiser’s state matches the state
they judge the Target to be in. The CARER task provides on- and off-line measures of each of these
processes, with the former relying on rapid intuitive judgement, and the latter on slower, reflective
processing thought to rely primarily on theory of mind (Schilbach, 2014; Schilbach et al., 2013).
Initial analyses focused on group comparisons with an individual differences approach to examine the
extent to which co-occurring alexithymia could account for any effect of autism (Bird & Viding,
2014; Bird & Cook, 2013).
Results showed that, after accounting for alexithymia, individuals with autism showed lower
offline emotion identification and affective empathy compared to non-autistic controls. However, we
found no evidence of reduced online emotion identification and affective empathy in our autistic
participants. Crucially, our findings also indicate that after controlling for alexithymia scores, autism
is not associated with affect sharing either online or offline.
It has previously been argued that an impairment in emotion identification will, in almost all
cases, lead to a deficit in affective empathy (Coll et al., 2017). Affective empathy is defined by the
degree to which the Empathiser’s state matches that of the Target; if the Empathiser does not know
which state the Target is in, then it is unlikely their state will match that of the Target. Coll and

Individuals with Autism Share Others’ Emotions:
Evidence from the Continuous Affective Rating and Empathic Responses (CARER) Task
23
colleagues argued that affect sharing is a more appropriate measure of the empathic process, as it
reflects the degree to which the Target’s state, as judged by the Empathiser, affects the Empathiser’s
state. It is interesting, then, that after controlling for alexithymia no deficit in affect sharing was
observed in the autism group.
What might be responsible for the less accurate emotion identification and reduced affective
empathy in autism? Emotion identification is a process that may rely on a multitude of processes
depending on the experimental stimuli and time available for processing. These processes include
rapid recognition of facial and vocal expressions of emotion, and slower incorporation of theory of
mind and understanding of the social situation in order to infer the Target’s emotion (Bird & Viding,
2014; Coll et al, 2017). Meta-analyses have suggested that emotion recognition is not consistently
impaired in autism (Harms et al., 2010), and indeed may be a product of co-occurring alexithymia
(Bothe et al., 2019; Brewer et al., 2016, Cook et al., 2013; Trevisan et al., 2016). Studies of theory of
mind in autism, however, consistently demonstrate that individuals with autism are impaired when
inferring the mental states of others (Baron-Cohen, 2000; Frith 1994, 2012). Thus, it is possible that
theory of mind impairments hinder the ability of individuals with autism to identify the emotion of
the Target, which, as described above, would result in reduced affective empathy but not impact
affect sharing. Such a possibility is consistent with the claim by Schilbach and colleagues (Schilbach,
2014; Schilbach et al., 2013) that offline processing – the only type of processing for which
individuals with autism were impaired – is characterised by an increased reliance on theory of mind.
The current study also found that, compared to non-autistic controls, individuals with autism
used reduced affective language when recording a video response for the Targets. This finding is in
line with previous research showing atypical affective language processing in autism (see review by
Lartseva, Dijkstra & Buitelaar, 2015). Such difficulties processing affective language have been
observed in studies of memory and attention. For example, individuals with ASD do not show a
memory advantage, consistently observed in neurotypicals, when processing emotional words
(Beversdorf et al., 1998; Gaigg and Bowler, 2008, 2009b). In the attention domain, affective

Individuals with Autism Share Others’ Emotions:
Evidence from the Continuous Affective Rating and Empathic Responses (CARER) Task
24
language processing in autism has been studied in the context of the ‘attentional blink’ effect. In the
attentional blink paradigm, the likelihood of participants detecting a target word depends on the time
elapsed between presentation of the previous target word (T1) and the current target word (T2). If by
the time T2 is presented, the participant’s attention is still engaged in T1, then they are less likely to
detect T2. However, affective words tend to override the attentional blink effect as they are detected
with high accuracy regardless of timing of presentation. Researchers have found that individuals with
autism do not show this advantage for emotional words and instead show no difference between
emotional and neutral words when presented as T2s (Corden et al., 2008; Gaigg and Bowler, 2009a).
Furthermore, a similar pattern of results has been found in neurotypical adults with high levels of
autistic traits in studies using emotional faces stimuli (English, Mybery & Visser, 2017; 2019).
Overall, research in both the memory and attention domains suggest that individuals with autism tend
to process both affective and neutral language similarly.
In the current study, when asked to record a message for the Target in the video clip,
participants are required to utilise theory of mind and their ability to produce spontaneous affective
language. Both, theory of mind (Baron-Cohen, 2000; Frith 1994, 2012) and spontaneous use of
affective language (Capps et al., 2000; Barnes et al., 2009) are known to be impaired in autistic
individuals. The current experimental design does not allow to speculate on the origin of
impoverished affective language in ASD; however, our data do show that the use of affective
language correlates with both AQ scores and, perhaps not surprisingly, with the TAS subscale
difficulty describing feelings. Future work could further examine the nature of these correlations with
affective language, for example by a) including both theory of mind and alexithymia measures; and
b) testing ASD participants with a wider range of alexithymia scores.
How do findings relate to current theories of empathy in autism?
On the whole, results did not support the emotional imbalance hypothesis advanced by Smith
(2009). This theory suggests that individuals with autism have a surfeit of affective empathy and are

Individuals with Autism Share Others’ Emotions:
Evidence from the Continuous Affective Rating and Empathic Responses (CARER) Task
25
more likely to experience empathic personal distress compared to neurotypical individuals. This
increase in personal distress is argued to lead to the behavioural differences in empathy frequently
reported in autistic individuals. However, the results of the current study revealed reduced affective
empathy in the autistic individuals, and evidence of reduced, rather than excessive, personal distress.
The reduction in personal distress was observed in two of the three measures; two measures derived
from the participants’ responses to the Target (verbal and non-verbal signs of personal distress)
showed reduced distress in the autistic group, while the autistic participants’ responses on the IRI
personal distress subscale were not significantly different to those of the neurotypical participants. It
should be noted, however, that the neurotypical raters may have misinterpreted or missed signs of
distress in the autistic individuals, as several studies suggest that neurotypical individuals are quite
poor at reading the emotional expressions of autistic individuals (e.g. Brewer et al., 2016; Volker et
al., 2009; Faso et al., 2015). In future, the inclusion of physiological measures of personal distress (or
at least arousal), such as heart rate, skin conductance or electromyography (EMG) recordings, could
be added to the CARER task to provide more objective evidence of personal distress.
Can alexithymia explain any empathic impairment in autism? The data support the co-
occurrence of alexithymia in ASD, as autistic participants scored significantly higher on the TAS
than the control group. Furthermore, alexithymia scores across all participants were negatively
correlated with affective empathy – both online and offline – and offline affect sharing. This means
that those with higher alexithymia scores reported emotional responses to the Target’s state that were
further away from both the Target’s actual state, and their estimate of the Target’s state. In addition,
controlling for alexithymia caused several apparent empathy-related deficits in autism to become
non-significant. Notably, alexithymia could not explain the reduced offline emotion identification and
affective empathy seen in the autism group. However, if this result is explained by poorer theory of
mind in the autism group as argued above, it would be consistent with work suggesting that theory of
mind is not impaired in alexithymia (Oakley, Brewer, Bird, & Catmur, 2016; Silani et al, 2008).

Individuals with Autism Share Others’ Emotions:
Evidence from the Continuous Affective Rating and Empathic Responses (CARER) Task
26
Nevertheless, it should be noted that the ability of alexithymia to explain CARER task
performance over and above autistic traits was limited to offline affect sharing. This may be a product
of a failure to match the autism and control groups for alexithymia (while 76% of participants in the
autism group scored above the alexithymia cut-off on the TAS, only 27% of the control group did). In
addition, the design of the CARER task mitigates one of the main difficulties in alexithymia;
identification of emotion. For both Self and Target ratings participants were required to judge the
intensity of emotion (from calm to upset) but not to determine whether the Target (or themselves)
was angry, afraid, or sad. Thus, it is likely that alexithymic difficulties would become more apparent
if the task required a judgement of what emotion is being felt, as well as the intensity of that emotion.
Thus, future work could investigate any alexithymia deficits in tasks requiring not only the intensity
of emotion to be judged, but also the type of emotion to be recognised.
Another important way in which the CARER task could be enhanced in future studies is by
adding positive emotions to investigate if the pattern of performance observed here is consistent with
empathy for positive emotions. The current study focused on negative emotions as this enabled
comparison with previous empathy research, which has overwhelmingly focused on negative
emotions. However, positive empathy is an emerging field (Morelli, Lieberman & Zaki, 2015), which
has the potential to contribute to current knowledge about the relevance of empathy for social
interactions and relationships in both neurotypical (Andreychik, 2017) and clinical populations
(Morrison et al., 2016). Finally, this work can also be extended by including other developmental
groups such as children and adolescents with and without autism. The inclusion of different age
groups could provide an insight into the developmental trajectory of empathic abilities and the
contribution of alexithymia to these abilities, both in neurotypical and autistic individuals.
The findings reported here imply that when appropriate measures are used, autistic
individuals do not show a lack of empathy. Note, however, that since we relied on self-reported ASD
trait measurements future work should seek to extend our results to examine if they hold when
additional measures of autism traits such as social skills, adaptive abilities, attention to detail, etc., are
assessed.

Individuals with Autism Share Others’ Emotions:
Evidence from the Continuous Affective Rating and Empathic Responses (CARER) Task
27
Concluding remarks
Contrary to previous work, our study, which employs a different methodological approach to
empathy research, shows that individuals with autism are not devoid of empathic abilities. Instead
they show less accurate retrospective emotion identification which, in turn, influences the extent to
which they report their affective state resembles that of the Target after a delay. Furthermore, there
was no effect of autism on affect sharing, implying that just like neurotypicals, individuals with
autism are able to share what they believe to be the emotions of others.

Individuals with Autism Share Others’ Emotions:
Evidence from the Continuous Affective Rating and Empathic Responses (CARER) Task
28
References
Adler, N., Dvash, J., & Shamay‐Tsoory, S. G. (2015). Empathic embarrassment accuracy in autism
spectrum disorder. Autism Research, 8(3), 241-249.
Andreychik, M. R. (2019). I like that you feel my pain, but I love that you feel my joy: Empathy for a
partner’s negative versus positive emotions independently affect relationship quality. Journal
of Social and Personal Relationships, 36(3), 834–854.
Avenanti, A., Bueti, D., Galati, G., & Aglioti, S. M. (2005). Transcranial magnetic stimulation
highlights the sensorimotor side of empathy for pain. Nature neuroscience, 8(7), 955.
Bagby, R. M., Taylor, G. J., & Parker, J. D. (1994). The twenty-item Toronto Alexithymia Scale—II.
Convergent, discriminant, and concurrent validity. Journal of psychosomatic research, 38(1),
33-40.
Baron-Cohen, S., Leslie, A., & Frith, U. (1985). Does the autistic child have a ‘theory of mind’?
Cognition, 21, 37–46.
Baron-Cohen, S. (2000). Theory of mind and autism: A fifteen year review. Understanding other
minds: Perspectives from developmental cognitive neuroscience, 2, 3-20.
Baron-Cohen, S., Richler, J., Bisarya, D., Gurunathan, N., & Wheelwright, S. (2003). The
systemizing quotient: an investigation of adults with Asperger syndrome or high–functioning
autism, and normal sex differences. Philosophical Transactions of the Royal Society B:
Biological Sciences, 358(1430), 361-374.
Baron-Cohen, S., & Wheelwright, S. (2004). The empathy quotient: an investigation of adults with
Asperger syndrome or high functioning autism, and normal sex differences. Journal of
autism and developmental disorders, 34(2), 163-175.
Baron-Cohen, S., Wheelwright, S., Skinner, R., Martin, J., & Clubley, E. (2001). The autism-
spectrum quotient (AQ): Evidence from asperger syndrome/high-functioning autism,

Individuals with Autism Share Others’ Emotions:
Evidence from the Continuous Affective Rating and Empathic Responses (CARER) Task
29
malesand females, scientists and mathematicians. Journal of autism and developmental
disorders, 31(1), 5-17.
Batson, C. D. (2009). These things called empathy: eight related but distinct phenomena.
Beck, A. T., Steer, R. A., & Brown, G. K. (1996). Manual for the beck depression inventory-II. San
Antonio, TX: Psychological Corporation, 1, 82.
Beversdorf, D. Q., Anderson, J. M., Manning, S. E., Anderson, S. L., Nordgren, R. E., Felopulos, G.
J., et al. (1998). The effect of semantic and emotional context on written recall for verbal
language in high functioning adults with autism spectrum disorder. J. Neurol Neurosurg.
Psychiatry 65, 685–692. doi: 10.1136/jnnp.65.5.685.
Bird, G., & Cook, R. (2013). Mixed emotions: the contribution of alexithymia to the emotional
symptoms of autism. Translational psychiatry, 3(7), e285.
Bird, G., Silani, G., Brindley, R., White, S., Frith, U., & Singer, T. (2010). Empathic brain responses
in insula are modulated by levels of alexithymia but not autism. Brain, 133(5), 1515-1525.
Bird, G., Press, C., & Richardson, D. C. (2011). The role of alexithymia in reduced eye-fixation in
autism spectrum conditions. Journal of autism and developmental disorders, 41(11), 1556-
1564.
Bird, G., Silani, G., Brindley, R., White, S., Frith, U., & Singer, T. (2010). Empathic brain responses
in insula are modulated by levels of alexithymia but not autism. Brain, 133(5), 1515-1525.
Bird, G., & Viding, E. (2014). The self to other model of empathy: providing a new framework for
understanding empathy impairments in psychopathy, autism, and alexithymia. Neuroscience
& Biobehavioral Reviews, 47, 520-532.
Bothe, E., Palermo, R., Rhodes, G. , Burton, N.; Jeffery, L. (2019). Expression Recognition
Difficulty Is Associated with Social But Not Attention-to-Detail Autistic Traits and Reflects
Both Alexithymia and Perceptual Difficulty. Journal of autism and developmental disorders,
1-13.

Individuals with Autism Share Others’ Emotions:
Evidence from the Continuous Affective Rating and Empathic Responses (CARER) Task
30
Brewer, R., Biotti, F., Catmur, C., Press, C., Happé, F., Cook, R., & Bird, G. (2016). Can
neurotypical individuals read autistic facial expressions? Atypical production of emotional
facial expressions in autism spectrum disorders. Autism Research, 9(2), 262-271.
Chevallier, C., Kohls, G., Troiani, V., Brodkin, E. S., & Schultz, R. T. (2012). The social motivation
theory of autism. Trends in cognitive sciences, 16(4), 231-239.
Chita-Tegmark, M. (2016). Social attention in ASD: a review and meta-analysis of eye-tracking
studies. Research in developmental disabilities, 48, 79-93.
Clements, C. C., Zoltowski, A. R., Yankowitz, L. D., Yerys, B. E., Schultz, R. T., & Herrington, J. D.
(2018). Evaluation of the social motivation hypothesis of autism: a systematic review and
meta-analysis. JAMA psychiatry, 75(8), 797-808.
Coll, M. P., Viding, E., Rütgen, M., Silani, G., Lamm, C., Catmur, C., & Bird, G. (2017). Are we
really measuring empathy? Proposal for a new measurement framework. Neuroscience &
Biobehavioral Reviews, 83, 132-139.
Cook, R., Brewer, R., Shah, P., & Bird, G. (2013). Alexithymia, not autism, predicts poor recognition
of emotional facial expressions. Psychological Science, 24(5), 723-732.
Corden, B., Chilvers, R., and Skuse, D. (2008). Emotional modulation of perception in Asperger’s
syndrome. J. Autism Dev. Disord. 38, 1072–1080. doi: 10.1007/s10803-007-0485-y.
Davis, M. H. (1980). A multidimensional approach to individual differences in empathy. JSAS
Catalogue of Selected Documents in Psychology, 10, 85.
Dawson, G., Webb, S. J., & McPartland, J. (2005). Understanding the nature of face processing
impairment in autism: insights from behavioral and electrophysiological studies.
Developmental neuropsychology, 27(3), 403-424.
De Vignemont, F., & Singer, T. (2006). The empathic brain: how, when and why? Trends in
cognitive sciences, 10(10), 435-441.

Individuals with Autism Share Others’ Emotions:
Evidence from the Continuous Affective Rating and Empathic Responses (CARER) Task
31
Decety, J., & Moriguchi, Y. (2007). The empathic brain and its dysfunction in psychiatric
populations: Implications for intervention across different clinical conditions.
BioPsychoSocial Medicine, 1(1), 22.
Dziobek, I., Rogers, K., Fleck, S., Bahnemann, M., Heekeren, H. R., Wolf, O. T., & Convit, A.
(2008). Dissociation of cognitive and emotional empathy in adults with Asperger syndrome
using the Multifaceted Empathy Test (MET). Journal of autism and developmental disorders,
38(3), 464-473.
English, M. C., Maybery, M. T., & Visser, T. A. (2017). Threatening faces fail to guide attention for
adults with autistic‐like traits. Autism Research, 10(2), 311-320.
English, M. C., Maybery, M. T., & Visser, T. A. (2019). Autistic-traits, not anxiety, modulate
implicit emotional guidance of attention in neurotypical adults. Scientific Reports, 9(1), 1-10.
Faso, D. J., Sasson, N. J., & Pinkham, A. E. (2015). Evaluating posed and evoked facial expressions
of emotion from adults with autism spectrum disorder. Journal of Autism and Developmental
Disorders, 1–15.
Frith, U. (1994). Autism and theory of mind in everyday life. Social development, 3(2), 108-124.
Frith, U. (2012). Why we need cognitive explanations of autism. The Quarterly Journal of
Experimental Psychology, 65(11), 2073-2092.
Gaigg, S. B., and Bowler, D. M. (2008). Free recall and forgetting of emotionally arousing words in
autism spectrum disorder. Neuropsychologia 46, 2336–2343. doi:
10.1016/j.neuropsychologia.2008.03.008.
Gaigg, S. B., and Bowler, D. M. (2009a). Brief report: attenuated emotional suppression of the
attentional blink in autism spectrum disorder: another non-social abnormality? J. Autism
Dev. Disord. 39, 1211–1217. doi: 10.1007/s10803-009-0719-2.

Individuals with Autism Share Others’ Emotions:
Evidence from the Continuous Affective Rating and Empathic Responses (CARER) Task
32
Gaigg, S. B., and Bowler, D. M. (2009b). Illusory memories of emotionally charged words in autism
spectrum disorder: further evidence for atypical emotion processing outside the social
domain. J. Autism Dev. Disord. 39, 1031–1038. doi: 10.1007/s10803-009-0710-y.
Harms, M. B., Martin, A., & Wallace, G. L. (2010). Facial emotion recognition in autism spectrum
disorders: A review of behavioral and neuroimaging studies. Neuropsychology Review, 20(3),
290–322.
Heaton, P., Reichenbacher, L., Sauter, D., Allen, R., Scott, S., & Hill, E. (2012). Measuring the
effects of alexithymia on perception of emotional vocalizations in autistic spectrum disorder
and typical development. Psychological medicine, 42(11), 2453-2459.
Hill, E., Berthoz, S., & Frith, U. (2004). Brief report: Cognitive processing of own emotions in
individuals with autistic spectrum disorder and in their relatives. Journal of autism and
developmental disorders, 34(2), 229-235.
Ickes, W., Stinson, L., Bissonnette, V., & Garcia, S. (1990). Naturalistic social cognition: Empathic
accuracy in mixed-sex dyads. Journal of Personality and Social Psychology, 59, 730– 742.
JASP Team (2019). JASP (Version 0.11.1)[Computer software].
Jeffreys, H. (1961). Theory of Probability (3rd Ed.). Oxford, UK: Oxford University Press.
Lartseva, A., Dijkstra, T., & Buitelaar, J. K. (2015). Emotional language processing in autism
spectrum disorders: a systematic review. Frontiers in human neuroscience, 8, 991.
Lawrence, E., Shaw, P., Baker, D., Baron-Cohen, S., & David, A. (2004). Measuring empathy:
Reliability and validity of the Empathy Quotient. Psychological Medicine, 34(5), 911-920.
doi:10.1017/S0033291703001624
Minio-Paluello, I., Baron-Cohen, S., Avenanti, A., Walsh, V., & Aglioti, S. M. (2009). Absence of
embodied empathy during pain observation in Asperger syndrome. Biological psychiatry,
65(1), 55-62.

Individuals with Autism Share Others’ Emotions:
Evidence from the Continuous Affective Rating and Empathic Responses (CARER) Task
33
Morelli, S. A., Lieberman, M. D., & Zaki, J. (2015). The emerging study of positive empathy. Social
and Personality Psychology Compass, 9(2), 57-68.
Morrison, A. S., Mateen, M. A., Brozovich, F. A., Zaki, J., Goldin, P. R., Heimberg, R. G., & Gross,
J. J. (2016). Empathy for positive and negative emotions in social anxiety disorder.
Behaviour research and therapy, 87, 232-242.
Raftery, A. E. (1995). Bayesian model selection in social research. Sociological methodology, 111-
163.
Reddy, V. (2003). On being the object of attention: implications for self-other consciousness. Trends
Cogn. Sci. 7, 397–402. doi: 10.1016/s1364-6613(03)00191-8
Rueda, P., Fernández-Berrocal, P., & Baron-Cohen, S. (2015). Dissociation between cognitive and
affective empathy in youth with Asperger Syndrome. European Journal of Developmental
Psychology, 12(1), 85-98.
Salminen, J. K., Saarijärvi, S., Äärelä, E., Toikka, T., and Kauhanen, J. (1999). Prevalence of
alexithymia and its association with sociodemographic variables in the general population of
Finland. J. Psychosom. Res. 1, 75–82. doi: 10.1016/S0022-3999(98)00053-1
Schilbach, L. (2014). On the relationship of online and offline social cognition. Frontiers in human
neuroscience, 8, 278.
Schilbach, L., Timmermans, B., Reddy, V., Costall, A., Bente, G., Schlicht, T., & Vogeley, K.
(2013). Toward a second-person neuroscience 1. Behavioral and brain sciences, 36(4), 393-
414.
Shamay-Tsoory, S. G., Tomer, R., Yaniv, S., & Aharon-Peretz, J. (2002). Empathy deficits in
Asperger syndrome: A cognitive profile. Neurocase, 8(2), 245-252.
Silani, G., Bird, G., Brindley, R., Singer, T., Frith, C., & Frith, U. (2008). Levels of emotional
awareness and autism: an fMRI study. Social neuroscience, 3(2), 97-112.

Individuals with Autism Share Others’ Emotions:
Evidence from the Continuous Affective Rating and Empathic Responses (CARER) Task
34
Smith, A. (2009). The empathy imbalance hypothesis of autism: a theoretical approach to cognitive
and emotional empathy in autistic development. The Psychological Record, 59(3), 489-510.
Trevisan, D. A., Bowering, M., & Birmingham, E. (2016). Alexithymia, but not autism spectrum
disorder, may be related to the production of emotional facial expressions. Molecular autism,
7(1), 46.
Volker, M. A., Lopata, C., Smith, D. A., & Thomeer, M. L. (2009). Facial encoding of children with
high-function autism spectrum disorders. Focus on Autism and Other Developmental
Disabilities, 24(4), 195–204.
Wechsler, D. (2009). Wechsler Memory Scale–Fourth Edition (WMS–IV) technical and interpretive
manual. San Antonio, TX: Pearson
Zaki, J., Bolger, N., & Ochsner, K. (2008). It takes two: The interpersonal nature of empathic
accuracy. Psychological science, 19(4), 399-40

Individuals with Autism Share Others’ Emotions:
Evidence from the Continuous Affective Rating and Empathic Responses (CARER) Task
35
Footnotes
1. Negative scores would indicate ‘excessive’ Affective Empathy, showing that the
Empathiser’s reaction to the Target’s affective state is stronger than the Target’s own
reaction.
2. When checking for homogeneity of variance using Levene’s test, the variable affective
emotion identification was significant at T2. Therefore, findings from the ANCOVA relating
to the crucial Group x Story type interaction were re-assessed using the non-parametric
Kruskal-Wallis test in JASP (JASP Team, 2019). We calculated a difference score between
affective and neutral trials to obtain a single difference score and compared this score
between the groups. The results of this analysis showed the same pattern of significance as
the ANCOVA results (X2(1) =.2, p = .655).
3. The Levene’s test for the variable emotion identification for affective stories was significant,
therefore, the Group x Story type interaction found in the ANCOVA was further assessed
with a non-parametric Kruskal-Wallis test. We calculated a difference scores between
affective and neutral stories to obtain a single difference score and compared this score
between the groups. This analysis supported the ANCOVA results, (X2(1) = 6.88, p = .009).