1252320

research-article2024
AUT0010.1177/13623613241252320Autism</italic>Cheang et al.

Special Issue: The Social Model in Autism Research

Autism

Do you feel me? Autism, empathic 1­–13

© The Author(s) 2024

accuracy and the double empathy problem Article reuse guidelines:

sagepub.com/journals-permissions
https://doi.org/10.1177/13623613241252320
DOI: 10.1177/13623613241252320
journals.sagepub.com/home/aut

Rachael TS Cheang , Maya Skjevling ,

Alexandra IF Blakemore, Veena Kumari and Ignazio Puzzo

Abstract
Empathy deficits in autism, particularly cognitive empathy, have been a long-held, but much debated assumption. An
alternative perspective challenging this deficit model is the ‘double empathy problem’, proposing that empathy difficulties
are bidirectional between autistic and non-autistic people. Despite this view gaining popularity, there has been limited
research examining whether non-autistic people can empathise accurately, cognitively and affectively with autistic
people. Addressing this gap, 81 adults from the general population, divided into groups based on how likely they are to
share personality traits common in autistic people, were examined using an empathic accuracy task, modified to include
autistic and non-autistic narrators and combined with a body mapping tool. Results showed participants had significantly
lower empathic accuracy scores when viewing autobiographical accounts of emotional events from autistic narrators,
compared to non-autistic narrators, especially for happy and sad emotions. However, participants also experienced
significantly higher intensity in the body when viewing autistic narrators compared to non-autistic narrators, especially
for anger and fear emotions. These findings support the double empathy problem and have strong implications for
therapeutic and interpersonal relationships with autistic people.

Lay Abstract
The assumption that autistic people lack empathy, particularly imagining how others feel, has been much debated and
is now being challenged by an alternative view: the ‘double empathy problem’. This suggests that non-autistic people
may find it equally difficult to imagine how autistic people feel. Although this perspective is gaining popularity, research
testing whether non-autistic people can accurately imagine and feel an autistic person’s emotions is still limited. Our
study used video clips of autistic and non-autistic people recounting emotional events to test if participants from the
general population could: track the intensity of the narrators’ emotions; name and feel the same emotion; match where
the narrator felt the emotion and indicate how intensely they felt the emotion using a body map. Our results show
that participants found it significantly harder to track autistic narrators’ emotions compared to non-autistic narrator’s
emotions, especially when viewing clips of narrators feeling happy and sad. We also found that participants felt emotions
more intensely in the body when viewing clips of autistic narrators compared to non-autistic narrators, especially
when describing anger and fear. These findings support the double empathy problem and have strong implications for
therapeutic and interpersonal relationships with autistic people.

Keywords
autism spectrum disorders, double empathy problem, empathy, ASD, empathic accuracy, autism

Autistic people have long been perceived by professionals

Brunel University London, UK
and society as having a lack of empathy (Baron-Cohen
et al., 2001). However, research has begun to critically Corresponding authors:
evaluate the term empathy when exploring autistic empa- Rachael TS Cheang, Department of Life Sciences, Brunel University
thy, as inconsistencies in defining terms often result in London, Quad North 134, Uxbridge, London UB8 3PH, UK.
confusion when interpreting research findings and more Email: [email protected]
importantly misunderstandings about autistic people within Ignazio Puzzo, Department of Life Sciences, Brunel University London,
the general population and among professionals (Blair, Quad North 134, Uxbridge, London UB8 3PH, UK.
2008; Bollen, 2023; Cuff et al., 2016; Fletcher-Watson & Email: [email protected]
2 Autism 00(0)

Bird, 2020; Harrison et al., 2022). For autistic people, such and AffEmp, showed autistic adults only demonstrated
misunderstandings can be painful and dehumanising, con- deficits in EmpAcc for anger compared to non-autistic
tributing to poor self-esteem and mental well-being (Cohen- adults. Furthermore, the autistic people studied showed no
Rottenberg, 2011; Nicolaidis et al., 2019). An alternative deficits in CogEmp or AffEmp on the EmpAcc task,
emerging view in autism research, challenging the notion despite reporting deficits in trait CogEmp and AffEmp on
of an empathy deficit altogether, proposes, instead, a ‘dou- a self-report measure (Mckenzie et al., 2022). However,
ble empathy problem’ (Milton et al., 2018). This suggests CogEmp and AffEmp on the EmpAcc task were measured
that when people with very different world views interact, by asking which emotion the participant thought the narra-
they struggle to empathise with each other, thus re-framing tor was feeling and how they felt watching the clip. This
the empathy issue as one of ‘reciprocity and mutuality’ yielded results that were either ‘right’ or ‘wrong’, generat-
(Milton, 2012). ing over-simplified measurements of AffEmp and
Empathy, defined broadly as ‘an emotional reaction in CogEmp. The authors noted this limitation suggesting that
an observer to the affective state of another individual’ more nuanced measures of these constructs were required
(Blair, 2005), is understood to be multidimensional, com- (Mckenzie et al., 2022). Santiesteban et al. (2021)
prising of at least two components: affective empathy described the CogEmp and AffEmp measures in the
(AffEmp) and cognitive empathy (CogEmp) (Davis, 1983; EmpAcc task as ‘offline’ measures (i.e. measured retro-
Dvash & Shamay-Tsoory, 2014; Eisenberg & Miller, 1987; spectively) and the EmpAcc measurement itself as an
Mazza et al., 2014; Zaki et al., 2008). While research sug- ‘online’ measure and so extended the EmpAcc task to
gests autistic people show differences in their ability to include continuous self-ratings. Results from this study
empathise (Baron-Cohen et al., 1997, 1985, 2001; Capps showed that autistic people only showed reduced CogEmp
et al., 1993; Happé, 1994; Harrison et al., 2022; Hill & and AffEmp compared to non-autistic people when meas-
Frith, 2003; Jones et al., 2010; Leppanen et al., 2018; ured offline, but not online; however, this task only asked
Magnée et al., 2007; Mazza et al., 2014; Rogers et al., participants to rate the intensity of emotion, rather than
2007), evidence that autistic people lack theory of mind identify which emotion was being felt, which in this study
(ToM), a component of CogEmp, appears more univer- was either sad or neutral only (Santiesteban et al., 2021).
sally accepted. A systematic review revealed that overall,
autistic people performed significantly less well than non-
The double empathy problem
autistic people, with large effect sizes on cognitive ToM
tasks (Leppanen et al., 2018). However, numerous empiri- The ‘double empathy problem’ was proposed based on
cal flaws have been highlighted in these studies Milton’s own experience, anecdotal accounts and limited
(Gernsbacher & Yergeau, 2019). Studies examining differ- amounts of qualitative data (Milton et al., 2022). However,
ences in AffEmp are more inconsistent: Results show an increasing number of studies have explored this con-
autistic people have similar (Mazza et al., 2014; Mul et al., cept. Crompton et al. (2020) showed that information
2018; Rogers et al., 2007), increased (Capps et al., 1993; passed through chains of people with similar neurotypes
Magnée et al., 2007) or decreased (Baron-Cohen & (autistic or non-autistic) was more accurate than through
Wheelwright, 2004; Lombardo et al., 2007) AffEmp com- chains of mixed neurotypes. Moreover, participant rapport
pared to non-autistic people. ratings were significantly lower for mixed chains.
While affective and cognitive components are often Similarly, studies report that autistic people feel less
measured concurrently in self-report questionnaires stressed interacting with other autistic people (Camus
(Baron-Cohen & Wheelwright, 2004; Davis, 1983), objec- et al., 2022), prefer within-neurotype interactions (Chen
tive performance tasks typically focus on only CogEmp or et al., 2021), create mutual understanding during interac-
AffEmp and use static images, failing to capture the more tions with other autistic people (Heasman & Gillespie,
nuanced, complex nature of empathy (Rum & Perry, 2019) and display fewer autistic traits when interacting
2020). Empathic accuracy (EmpAcc) tasks attempt to with other autistic people (Gernsbacher et al., 2017).
address this by measuring the ability to accurately infer Furthermore, interpersonal similarity of autistic traits
thoughts and feelings of another using video clips, which within the general population is associated with increased
are more dynamic and ecologically valid (Ickes et al., friendship quality (Bolis et al., 2021).
1990; Mackes et al., 2018; Mckenzie et al., 2022; Zaki There has been some exploration of how well non-
et al., 2008). Synthesised findings from a scoping review autistic people interpret the emotional expressions and
revealed autistic people and non-autistic people with per- mental states of autistic people. Brewer et al. (2016)
sonality traits similar to those assumed common among reported non-autistic observers as less able to recognise
the autistic community, exhibit deficits in EmpAcc (Rum photographed autistic emotional expressions compared to
& Perry, 2020). non-autistic expressions. Similarly, they found autistic
Nonetheless, a recent study employing a modified observers equally poor at recognising autistic emotional
EmpAcc task, including additional measures of CogEmp expressions. Sheppard et al. (2016) conducted a series of
Cheang et al. 3

studies where non-autistic participants were shown short, that participants from the general population would strug-
muted videos in which autistic and non-autistic targets gle to empathise accurately, cognitively and affectively,
responded to one of four scenarios (joke, waiting, story with the emotions of autistic narrators. Second, we aimed
and compliments). Participants had to decide which sce- to test whether individuals from the general population
nario provoked which response, score how expressive tar- with Autism-Spectrum Quotient (AQ) scores closer to
gets were and describe the response. Although participants those found in autistic people empathise more accurately,
found autistic targets as expressive as non-autistic targets cognitively and affectively with autistic narrators than
in 75% of scenarios, they found it significantly more dif- with non-autistic narrators compared to those with a
ficult to identify which scenario autistic targets were greater mismatch in AQ scores. Although the AQ has been
responding to compared to non-autistic targets (Sheppard used previously as a measure of ‘autistic traits’ in the gen-
et al., 2016), perhaps demonstrating their struggle to repre- eral population, it should be noted that autistic traits are
sent mental states of autistic individuals. Edey et al. (2016) not homogeneous among autistic people and assumptions
used animations of triangles generated by autistic and non- of what constitutes low, medium, or high autistic traits are
autistic people to depict mental state interactions (coaxing, subjective judgements. Nonetheless, given that there is
mocking, seducing and surprising) seeing if autistic and some evidence to suggest that higher interpersonal similar-
non-autistic people could infer correct mental states from ity of autistic traits in the general population is associated
both groups. Researchers found that non-autistic perceiv- with higher measures of closeness, acceptance and help
ers had enhanced ability to infer correct mental states on (Bolis et al., 2021), we predicted that participants from the
non-autistic animations compared to autistic animations, general population with higher AQ scores would empa-
but autistic perceivers showed no difference in identifying thise more accurately, cognitively and affectively with
mental states of animations created by autistic and non- autistic narrators compared to non-autistic narrators than
autistic participants. non-autistic participants with lower AQ scores. Our novel
Although the above studies show that (1) non-autistic approach combined a modified version of the EmpAcc
people struggle to represent mental states of autistic people task (N. A. Martin-Key et al., 2017) manipulating the nar-
and (2) non-autistic people struggle to recognise autistic rator type to include autistic and non-autistic narrators
people’s emotions from photographed facial expressions, with a modified version of the emBODY tool (Nummenmaa
the stimuli used did not adequately capture the dynamic, et al., 2014), which maps bodily sensations associated with
nuanced and complex expression of emotion. Furthermore, emotions, obtaining a more refined measure of AffEmp.
these studies only focused on CogEmp and not AffEmp, an The emBODY tool has been used in previous research to
embodied empathic response to another’s emotions. The examine how body maps of emotion are related to trait
present study aims to fill these gaps, undoubtedly impor- measures of empathy (Sachs et al., 2019).
tant in therapeutic and interpersonal relationships. Camm-
Crosbie et al. (2019) developed an online survey exploring
experiences of autistic adults receiving treatment and sup- Method
port for mental health problems. Three underlying themes
emerged: (1) many autistic adults, perceived as ‘coping’,
Participants
were dismissed for treatment; (2) many professionals Eighty-five adults from the general population were
lacked understanding of how autistic people communicate, recruited through the Brunel University Intranet and vari-
socially interact and crucially show and express feelings ous social media platforms. An a priori power analysis
and emotions; and (3) lack of appropriate support con­ using G-Power suggested a sample size of 87 based on a
tributed to poor mental health outcomes, self-injury and repeated measures analysis of variance (ANOVA) with
suicidality in autistic people. It is, therefore, crucial to three groups and 10 performance measures, detecting an
understand whether non-autistic people can accurately effect size of 0.15 and significance level of 0.05% and 95%
empathise with autistic people’s emotions. power. Participants were eligible for the study if they could
travel to Brunel University London for the study; were
aged 18–85 years; were fluent in English; had no known
The current study history of psychosis, substance abuse or a traumatic brain
Primarily, this study aimed to understand whether people injury; and had no genetic disorders affecting brain func-
from the general population (not considered autistic), can tion or intellectual disability. The participants did not dis-
accurately, cognitively and affectively empathise with close whether they were autistic or not, but the mean AQ
autistic and non-autistic narrator’s emotions. Based on score fell below the clinical cut off and so participants were
previous research showing non-autistic people struggling considered non-autistic. Ethical approval was granted by
to represent mental states of autistic people and that autis- the College of Health, Medicine and Life Sciences Research
tic people feel their emotions are misunderstood (Camm- Ethics Committee, Brunel University London (reference:
Crosbie et al., 2019; Sheppard et al., 2016), we expected 32839-A-May/2022-39560-2). All participants provided
4 Autism 00(0)

written consent and received £10 (Amazon voucher) or

course credits for their participation.

Measure of autistic traits

Autistic traits in our sample from the general population
were measured using the 50-item self-report AQ (Baron-
Cohen et al., 2001). The scale has good psychometric
properties, with normally distributed AQ sum scores (score
range 0–50, with ⩾29 indicating clinically significant
autism traits) in the general population (Hurst et al., 2007).
The scale had a high level of internal consistency in the
current sample (Cronbach’s α = 0.848).

Behavioural tasks
The empathic accuracy task. The EmpAcc task was first
Figure 1. Example of continuous rating during the empathic
designed to measure perceivers’ ability to accurately accuracy task.
assess targets’ emotions (Zaki et al., 2008) and was subse- Note. Participants were asked to rate on a 9-point scale how emotional
quently modified by Martin-Key et al. (2017) to include they thought the narrator was feeling while recounting an autobiographical
videos depicting primary emotions instead of undifferen- event. 0 indicated no emotion, 9 indicated strong emotion.
tiated positively and negatively valenced stimuli. We fur-
ther modified the task to include autistic narrators and
non-autistic narrators to allow the double empathy prob-
lem to be examined (description of task construction in
Supplementary Material). During the EmpAcc task, par-
ticipants were shown 10 pseudo-randomised video clips
of narrators giving autobiographical accounts of situa-
tions where they felt one of four primary emotions (anger,
fear, happiness and sadness) and a neutral event. The par-
ticipants were instructed to continually rate the intensity
of the emotion on the 9-point scale while viewing a train-
ing clip (see Figure 1). It was emphasised that the partici-
pant should rate the intensity of the emotion being
experienced by the narrator while they were speaking
about the event, not the intensity felt during the event
itself. An EmpAcc score was formed by correlating the
participants’ continuous ratings of emotional intensity to
the narrators’. The participants were then asked to name Figure 2. Example of body mapping tool.
the emotion the narrator was experiencing while speaking Note. After watching each video clip, participants were asked to identify
(options: happy, sad, frightened, angry, surprised and no on the body map where they felt sensation and at what intensity on a
10-point scale. 0 indicated no sensation, 10 indicated intense sensation.
emotion) (CogEmp), as well as the emotion they felt dur-
ing the task (same six options) (AffEmp).
calculated (BodyEmp); a match in a body part was given a
The emBODY task. The emBODY task is a computer- score of 1 if the participant and the narrator both had a felt
based, topographical self-report task developed to map sense of any intensity or both had no felt sense in that body
bodily sensations associated with different emotions part. A maximum score of 15 meant a total match. The
(Nummenmaa et al., 2014). Different emotions were con- participants’ average intensity rating across all body parts
sistently associated with statistically separable bodily sen- (ParInt) was also calculated.
sation maps. The present study modified the maps so that
15 discrete areas of the body (top head, eyes, cheeks, jaw,
Procedure
throat, shoulders, arms, hands, chest, heart, stomach, abdo-
men, pelvis, legs and feet) could be rated for intensity The participants were given a participant information sheet
of feeling on a scale of 0–10, 0 being no sensation and stating the aims of the research and a consent form to com-
10 being intense sensation (see Figure 2). The total number plete. Demographic information (age, gender and ethnic ori-
of body parts that matched the narrators’ body parts was gin) was collected; however, data on socioeconomic status
Cheang et al. 5

and educational attainment levels were not recorded. The transformation of the data: The median version of Levene’s
participants completed the AQ self-report measure, fol- test for equality of variances for use, as suggested by
lowed by the two behavioural tasks. The participants were Brown and Forsythe (1974), was homogeneity of vari-
unaware of the diagnostic status of the narrators during the ances (p > 0.05). There was homogeneity of covariances,
task but were debriefed after completion of the study. as assessed by Box’s test of equality of covariance matri-
ces: for CogEmp p = 0.305 and AffEmp p = 0.372.
Analysis
Community involvement
Collected data were analysed using SPSS v.29 software.
The mean AQ score was calculated, and the sample split The research team included autistic (first author) and non-
into three groups using quartiles: Low-AQ (lower quartile, autistic researchers. The study design and research questions
AQ < 13), High-AQ (upper quartile, AQ > 24) and were formulated by the autistic researcher. Furthermore, the
Medium-AQ (interquartile range, AQ = 13–24). Four par- video clips used in the EmpAcc task were filmed with two
ticipants were excluded as they completed <75% of the autistic and two non-autistic narrators. There was no further
behavioural task, leaving 81 participants (21 in Low-AQ community involvement in the design of this study.
group, 38 in Medium-AQ group, and 22 in High-AQ
group). Possible group differences in demographic data Results
were examined using a one-way ANOVA (age) or chi-
square tests (gender and ethnicity). Demographics
To test the double empathy hypothesis, three-way mixed
There were no Group differences in age, F(2,77) = 1.61,
ANOVAs with one between-subjects factor (Group:
p = 0.207; gender, χ2(4) = 4.12, p = 0.390; or ethnicity,
Low-AQ/Medium-AQ/High-AQ) and two within-subject
χ2(12) = 9.85, p = 0.629 (Table 1).
factors (Narrator-Type: autistic/non-autistic, Emotion:
angry/sadness/fear/happiness/neutral) were run to deter-
mine the effect of different Narrator-Types and different Empathic accuracy (EmpAcc)
Emotions on EmpAcc, BodyEmp and ParInt between par-
There was a significant main effect of Narrator-Type, F(1,
ticipants with High-AQ, Medium-AQ and Low-AQ. There
73) = 8.95, p = 0.004, partial η2 = 0.109, with participants
were multiple outliers in the data, as assessed by inspection
across the sample showing lower EmpAcc when viewing
of a boxplot, but were kept in the analysis because they did
clips of autistic narrators than clips of non-autistic narra-
not materially affect the results as assessed by a comparison
tors. There was a significant two-way interaction between
of the results with and without the outliers. EmpAcc,
Narrator-Type and Emotion, F(2.85, 208.17) = 4.592,
BodyEmp and ParInt scores were not normally distributed,
p = 0.005, ε = 0.713, partial η2 = 0.059 (see Table 2 and
as assessed by Shapiro-Wilk’s test (p > 0.05). As ANOVAs
Figure 3). Therefore, simple main effects were run with
are fairly robust against deviations from normality and the
Bonferroni adjustment applied. EmpAcc was significantly
sample size was fairly large, we continued without trans­
different for non-autistic narrators (M = 0.91, SD = 0.07)
formation of the data. However, as the underlying data had
compared to autistic narrators (M = 0.86, SD = 0.12) for
a skewed distribution, we used the median version of
sadness, F(1, 73) = 7.77, p = 0.007, partial η2 = 0.096, a
Levene’s test for equality of variances as suggested by
mean difference of –0.043, 95% confidence interval (CI)
Brown and Forsythe (1974), which revealed homogeneity
= [–0.074, –0.012] and for non-autistic narrators (M = 0.87,
of variances (p > 0.05). For the three-way interaction
SD = 0.13) compared to autistic narrators (M = 0.78,
effect, Mauchly’s test of sphericity indicated that the
SD = 0.21) for happiness, F(1, 73) = 9.28, p = 0.003, partial
assumption of sphericity had been violated for EmpAcc,
η2 = 0.113, a mean difference of –0.09, 95% CI = [–0.15,
χ2(9) = 52.10, p < 0.001; BodyEmp, χ2(9) = 136.38,
–0.03]. EmpAcc was not significantly different for non-
p < 0.001; and ParInt, χ2(9) = 65.01, p < 0.001. The esti-
autistic narrators compared to autistic narrators for anger,
mated epsilons (ε) were 0.713, 0.524 and 0.753, respec-
fear or the neutral condition. There was no significant
tively; therefore, as suggested by Maxwell and Delaney
three-way interaction between Narrator-Type, Emotion
(2004), the Greenhouse-Geisser correction was used.
and Group. There were no significant two-way interactions
Two-way mixed ANOVAs with one between-subjects
between Narrator-Type and Group or Emotion and Group
factor and one within-subjects factor were also performed
(see Supplementary Material for non-significant results).
to determine the effect of different Narrator-Types on
CogEmp and AffEmp, between participants with High-AQ,
Medium-AQ and Low-AQ. There were no outliers, as Participant interoceptive intensity
assessed by examination of studentised residuals for val-
(ParInt)
ues greater than ±3. CogEmp and AffEmp scores were not
normally distributed, as assessed by Shapiro-Wilk’s test There was a significant main effect of Narrator-Type, F(1,
(p > 0.05). As above, we decided to continue without 78) = 17.19, p < 0.001, partial η2 = 0.181, with participants
6 Autism 00(0)

Table 1. Demographic characteristics of the sample, classified by group.

Low-AQ < 13 (n = 21a) Medium-AQ = 13–24 (n = 38) High-AQ > 24 (n = 22) p value
Age 32.6 (13.7) 27.3 (10.5) 32.2 (15.1) 0.207
Gender 0.390
Female 12 (22.2) 27 (50) 15 (27.8)
Male 9 (34.6) 11 (42.3) 6 (23.1)
Other 0 (0) 0 (0) 1 (100)
Ethnicity 0.629
Arab 1 (16.7) 3 (50) 2 (33.3)
Asian 5 (25) 11 (55) 4 (20)
Black 0 (0) 3 (75) 1 (25)
Latina 0 (0) 1 (100) 0 (0)
Mixed 3 (60) 1 (20) 1 (20)
Persian 0 (0) 0 (0) 1 (100)
White 12 (27.3) 19 (43.2) 13 (29.5)
AQ score 9.1 (2.1) 18.9 (3.7) 28.9 (3.2)

Note. Mean (std. dev) is given for age, all other values show numbers of participants (%).
a
n = 20 for age only as data for one participant was missing.

across the sample showing higher ParInt when viewing Cognitive empathy (CogEmp)
clips of autistic narrators than clips of non-autistic narra-
tors. There was a significant two-way interaction between The main effect of Group was approaching significance,
Narrator-Type and Emotion, F(3.01, 234.86) = 3.576, F(2, 78) = 2.97, p = 0.057, partial η2 = 0.07 (High-AQ
p = 0.002, ε = 0.753, partial η2 = 0.044 (see Table 2 and Group, M = 58.2, SD = 17.9; Medium-AQ Group, M = 66.1,
Figure 4). Therefore, simple main effects were run with SD = 12.0; Low-AQ Group, M = 68.6, SD = 16.2) (see
Bonferroni adjustment applied. ParInt was significantly Table 2 and Figure 5). Post hoc analysis with a Bonferroni
lower for non-autistic narrators (M = 0.59, SD = 0.98) adjustment revealed that the mean difference in CogEmp
compared to autistic narrators (M = 0.91, SD = 1.25) for between participants in the Low-AQ Group and the
anger, F(1, 78) = 21.19, p < 0.001, partial η2 = 0.214, a High-AQ Group (10.39, 95% CI = [–0.73, 21.50]) was
mean difference of 0.33, 95% CI = [0.19, 0.47] and for approaching statistical significance, p = 0.075. The main
non-autistic narrators (M = 0.55, SD = 1.08) compared to effect of Narrator-Type was not significantly different for
autistic narrators (M = 0.95, SD = 1.40) for fear, F(1, CogEmp, F(1, 78) = 0.17, p = 0.685, partial η2 = 0.002.
78) = 11.814, p < 0.001, partial η2 = 0.132, a mean differ- Participants across the sample did not differ in CogEmp
ence of 0.40, 95% CI = [0.17, 0.64]. ParInt was not sig- when viewing clips of autistic narrators compared to clips
nificantly different for non-autistic narrators compared to of non-autistic narrators. There was no significant interac-
autistic narrators for sadness, happiness or the neutral con- tion between Narrator-Type and Group, F(2, 78) = 0.27,
dition. There was no significant three-way interaction p = 0.764, partial η2 = 0.007 (see Table 2).
between Narrator-Type, Emotion and Group. There were
no significant two-way interactions between Narrator- Affective empathy (AffEmp)
Type and Group or Emotion and Group (see Supplementary
Material for non-significant results). There was a significant difference in AffEmp between
participants with High-AQ (M = 40.5, SD = 13.3),
Medium-AQ (M = 52.1, SD = 16.5) and Low-AQ (M = 51.4,
SD = 14.9), F(2, 78) = 4.49, p = 0.014, partial η2 = 0.103
Interoceptive empathy (BodyEmp)
(see Table 2 and Figure 6). Post hoc analysis with a
There was no significant main effect of Narrator-Type; Bonferroni adjustment revealed that the mean difference
participants across the sample did not differ in BodyEmp in AffEmp between participants in the Medium-AQ Group
when viewing clips of autistic and non-autistic narrators. and the High-AQ Group (11.65, 95% CI = [1.64, 21.66])
There was no significant three-way interaction involving was significant (p = 0.017) and the mean difference in
Narrator-Type, Emotion and Group. There were also no AffEmp between participants in the Low-AQ Group and
significant two-way interactions between Narrator-Type the High-AQ Group (10.97, 95% CI = [–0.42, 22.37]) was
and Group, Emotion and Group, or Narrator-Type and approaching statistical significance, p = 0.063. The main
Emotion (see Supplementary Material for non-significant effect of Narrator-Type was not significant for AffEmp,
results). F(1, 78) = 0.16, p = 0.687, partial η2 = 0.002. There was no
Cheang et al. 7

Table 2. ANOVA statistics for behavioural task performance.

Variable dfNum dfDen ε F p η2

EmpAcc
Narrator 1 73 8.95** 0.004 0.109
Emotion 2.76 201.18 0.689 9.61*** <0.001 0.116
Group 2 73 0.21 0.811 0.006
Narrator × Group 2 73 0.713 0.21 0.808 0.006
Emotion × Group 5.51 201.18 0.689 0.20 0.971 0.005
Narrator × Emotion 2.85 208.17 0.713 4.59** 0.005 0.059
Narrator × Emotion × Group 5.70 208.17 0.713 0.61 0.717 0.016
BodyEmp
Narrator 1 78 0.64 0.428 0.008
Emotion 2.06 160.47 0.514 5.51** 0.004 0.066
Group 2 78 0.02 0.985 0.000
Narrator × Group 2 78 0.524 0.78 0.462 0.020
Emotion × Group 4.12 160.47 0.514 0.21 0.937 0.005
Narrator × Emotion 2.10 163.45 0.524 0.39 0.690 0.005
Narrator × Emotion × Group 4.11 163.45 0.524 0.90 0.469 0.023
ParInt
Narrator 1 78 17.19*** <0.001 0.181
Emotion 3.28 255.72 0.820 16.96*** <0.001 0.179
Group 2 78 1.958 0.148 0.048
Narrator × Group 2 78 0.753 1.088 0.342 0.027
Emotion × Group 6.56 255.72 0.820 1.114 0.355 0.028
Narrator × Emotion 3.01 234.86 0.753 3.576* 0.015 0.044
Narrator × Emotion × Group 6.02 234.86 0.753 0.478 0.825 0.012
CogEmp
Narrator 1 78 0.17 0.685 0.002
Group 2 78 2.97 0.057 0.071
Narrator × Group 2 78 0.27 0.764 0.007
AffEmp
Narrator 1 78 0.16 0.687 0.002
Group 2 78 4.49* 0.014 0.103
Narrator × Group 2 78 0.14 0.869 0.004

Note. dfNum indicates degrees of freedom numerator. dfDen indicates degrees of freedom denominator. ε indicates Greenhouse-Geisser multiplier for
degrees of freedom, p values and degrees of freedom incorporate this correction.
*p < 0.05. **p < 0.01. ***p < 0.001.

significant interaction between Narrator-Type and Group, describing happy or sad events. Inversely, non-autistic par-
F(2, 78) = 0.14, p = 0.869, partial η2 = 0.004 (see Table 2). ticipants felt the emotions of autistic narrators significantly
more intensely in the body compared to non-autistic narra-
tors and that this effect was primarily driven by narrators
Discussion describing events where they felt anger or fear. Despite
This novel study investigated non-autistic participants’ emotions being felt more intensely in the body, the loca-
ability to continuously track changes in emotional inten- tions of where non-autistic participants and narrators felt
sity (EmpAcc) in autistic and non-autistic narrators while the emotions matched, regardless of whether narrators
recounting autobiographical emotional events. Participants were autistic or not.
were asked to identify the emotion they thought narrators Although no previous studies have measured EmpAcc,
were feeling (CogEmp), label the emotion they felt while ParInt or BodyEmp in non-autistic participants while
watching the clips (AffEmp) and indicate on a body map viewing clips of autistic and non-autistic narrators, these
where they were feeling the emotion (BodyEmp) and at findings are comparable with those reported by Brewer
what intensity they felt the emotion (ParInt). et al. (2016), who found non-autistic participants struggle
Consistent with our hypothesis, we found our sample of to recognise emotions of autistic participants from photo-
non-autistic participants had significantly less EmpAcc graphed facial expressions; by Sheppard et al. (2016) and
with the emotions of autistic narrators compared to non- Edey et al. (2016) who found non-autistic participants
autistic narrators – an effect primarily driven by narrators struggle to represent mental states of autistic compared to
8 Autism 00(0)

Figure 6. Affective empathy score by group.

Figure 3. Mean empathic accuracy scores by narrator-type
Note. Affective empathy scores are calculated as a percentage of
and emotion. correctly identified emotions (error bars show 95% confidence
Note. Mean empathic accuracy scores by Narrator-Type and Emption interval).
(error bars show 95% confidence interval).
*p < 0.01.
One possible explanation for this finding could be that
autistic individuals express emotion differently from non-
autistic individuals. Zaki et al. (2008) found expressivity
of the narrator a significant predictor of EmpAcc. Although
early accounts of autism describe autistic individuals as
having flat affect (Kanner, 1943), recent research suggests
they express emotions atypically (Brewer et al., 2016;
Faso et al., 2015; Volker et al., 2009). Some research dem-
onstrates emotion recognition was linked to atypical facial
and vocal expression (Brewer et al., 2016; Macdonald
et al., 1989), whereas other research showed facial and
vocal expressiveness was more awkward in autistic indi-
viduals, yet expressions were as recognisable as non-autis-
tic individuals’ expressions (Grossman et al., 2013).
Figure 4. Mean participant intensity score by narrator-type Moreover, Sheppard et al. (2016) found that although non-
and emotion.
Note. Mean participant intensity scores by Narrator-Type and Emotion
autistic participants struggled to assign correct mental
(error bars show 95% confidence interval). states to autistic targets, they were rated as similarly
*p < 0.001. expressive to the non-autistic targets. Inconsistencies in
these findings have been linked to differences in sex, age,
intelligence and context, which all play a role in how
expressive autistic samples appear (Begeer et al., 2008;
Capps et al., 1993; Faso et al., 2015). For example, Capps
et al. (1993) found autistic children with intellectual disa-
bility less expressive than those without and Faso et al.
(2015) found emotional expressions produced by autistic
females were less intense and less identifiable, but more
natural than those by autistic males.
Another possibility for inconsistency could be the pres-
ence or absence of alexithymia, which co-occurs in 50% of
autistic people (Kinnaird et al., 2019). One study found
judges were less able to decode spontaneous facial expres-
sions of participants with alexithymia compared to those
Figure 5. Cognitive empathy score by group.
Note. Cognitive empathy scores are calculated as a percentage of without alexithymia and rated their expressions as less
correctly identified emotions (error bars show 95% confidence intense (McDonald & Prkachin, 1990). However, when
interval). the same participants were asked to pose facial emotions,
alexithymic participants’ and non-alexithymic partici-
non-autistic individuals; and by Faso et al. (2015) who pants’ expressions were judged as accurately, but alex-
found autistic expressions were rated as more intense than ithymic participants’ expressions were significantly less
non-autistic expressions. intense than the non-alexithymic participants’ expressions
Cheang et al. 9

for anger and happiness only. This finding may suggest negatively impact social interaction. Being misinterpreted
that alexithymic people can force expression, but it comes encourages autistic people to camouflage to feel socially
across as less natural and, therefore, less intense. accepted, which can lead to burnout (Grace et al., 2022;
Conversely, Faso et al. (2015) reported that autistic expres- Raymaker et al., 2020). Moreover, feeling unsupported
sions were judged more accurately than non-autistic when distressed can lead to feelings of isolation. Loneliness
expressions, during both posed and evoked expressions of has been shown to contribute to low self-esteem and
anger, but less accurately than non-autistic expressions extremely high levels of depression, self-harm and suici-
during posed expressions of happiness. Furthermore, dality in autistic people (Grace et al., 2022; Hedley et al.,
aligned with our findings, autistic expressions were rated 2018; Mazurek, 2014; Moseley et al., 2021).
as significantly more intense than non-autistic expressions We further explored whether participants with higher
during the evoked condition for anger and happiness and AQ scores would empathise more accurately, cognitively
the posed condition for anger and fear, but significantly and affectively with autistic narrators compared to non-
less intense for sadness in the evoked condition only. Faso autistic narrators than participants with lower AQ scores.
et al. (2015) did not check whether autistic participants We did not find any significant interactions between
also had alexithymia. Future research could examine Narrator-Type and Group for EmpAcc, BodyEmp, ParInt,
whether alexithymia is a significant predictor of emotion CogEmp or AffEmp, meaning participants from all three
expressivity in autistic people. groups empathised as accurately and equally well, cogni-
Our data similarly showed the main effect of Narrator tively and affectively, with autistic compared to non-autis-
on EmpAcc was driven by specific emotions, namely, hap- tic narrators. Although we hypothesised that having higher
piness and sadness, but not anger or fear. This is consistent similarity of autistic traits might result in greater empathy,
with previous findings; for example, Volker et al. (2009) the participants in our High-AQ Group had a mean AQ of
reported that non-autistic raters were significantly less 28.9, whereas the mean AQ score in the clinical autistic
able to correctly identify the target emotion as happy or population is 35.2 (Ruzich et al., 2015). Furthermore, the
sad for autistic children compared to non-autistic children, AQ is a self-report measure and although it may be used as
although when controlling for IQ, only the difference a screening tool, scoring highly on this measure is not the
between groups for sadness remained. Similarly, both Faso same as being diagnosed as autistic; for example, people
et al. (2015) and Brewer et al. (2016) found happiness the with other conditions such as anxiety and schizophrenia
hardest emotional expression to recognise in autistic adults may have high levels of autistic traits (Sasson & Bottema-
compared to non-autistic adults. Yet, our finding also Beutel, 2022). It must also be emphasised that none of our
shows that while EmpAcc scores were similar for autistic participants were asked if they had an autistic diagnosis:
and non-autistic narrators for anger and fear, autistic narra- This was a clear limitation in our study design and future
tors evoked more intense feelings in the body for these studies should include diagnosed autistic participants.
emotions than non-autistic narrators. Feeling anger and Contrary to our prediction, our findings showed non-
fear more intensely in the body could have led participants autistic participants having equal ability to name (CogEmp)
to track the autistic narrators’ emotions more accurately. and share (AffEmp) the emotions of autistic and non-autis-
Sachs et al. (2019), for example, found a correlation tic narrators. This could be because participants were only
between trait perspective taking scores and the degree of given six options and could have guessed the narrator’s
overlap between self and other body maps. emotion. If guessed correctly, the participant may have felt
These results have important implications for both ther- social pressure to empathise with the narrator, despite not
apeutic and interpersonal relationships. For example, non- actually feeling the same way. Displaying empathic behav-
autistic people finding it difficult to understand when an iour is socially desirable and contributes to response bias
autistic person is happy, may fail to celebrate life’s joys in self-reported empathy measures (Sassenrath, 2020).
with them. Equally, if non-autistic people cannot recognise Although no main effect of Narrator-Type was found
when an autistic person is upset, they will fail to offer for CogEmp or AffEmp, we did find main effects of Group.
appropriate comfort and support. The lack of sharing in Participants in the High-AQ Group had lower mean
one’s happiness and comforting in one’s sadness may lead CogEmp and AffEmp scores than those in the Low-AQ
to autistic people to feel isolated, frustrated and angry. Group and significantly lower mean AffEmp scores than
Furthermore, autistic people’s anger and fear being felt those in the Medium-AQ Group. Interestingly, these find-
more intensely during interactions may lead to autistic ings were only evident in the empathic measures requiring
people being misperceived as ‘less warm’ or perceived the participant to label the emotion, whereas both the
‘less favourably’ than non-autistic people, as found in EmpAcc score measured during the EmpAcc task and the
studies examining real-world social interactions between ParInt score measured during the modified emBODY task
autistic and non-autistic adults (Morrison et al., 2020; rated emotional intensity rather than an ability to give
Sasson et al., 2017). Faso et al. (2015) also posited that emotions perceived labels. Perhaps this is again due to co-
misinterpreting autistic people as angrier than they are can occurring alexithymia and difficulty labelling emotions.
10 Autism 00(0)

Santiesteban et al. (2021) reported autistic participants and fear. This is a novel finding which supports the double
using reduced affective language compared to non-autistic empathy hypothesis and has strong implications for both
controls. They also found autistic participants exhibiting therapeutic and interpersonal relationships with autistic
lower empathy scores compared to non-autistic controls people. Failure to empathise with autistic people’s emo-
when participants were asked to rate narrators’ affects ret- tions could have detrimental effects on their self-esteem,
rospectively (offline). Conversely, when participants were mental health and well-being and how well they are sup-
asked to rate the affective states of narrators continuously ported. It is, therefore, crucial that awareness of differ-
while watching the clips (online), they reported no differ- ences in how autistic people communicate and express
ences between autistic and non-autistic participants’ empa- emotion is emphasised in training of caregivers, educators,
thy ratings. In contrast to these findings, McKenzie et al. healthcare practitioners and therapists. Future research
(2022) found no differences in total CogEmp and AffEmp should concentrate on how to increase empathy among the
between autistic and non-autistic participants during the normative population and reduce the responsibility and
EmpAcc task; however, this study lacked statistical power burden of autistic people having to fit in.
and so may have failed to detect group differences.
Authors’ note
Strengths and limitations I am an autistic White female researcher diagnosed at the age of
30. Being undiagnosed for the majority of my life, I have strug-
This study had several strengths. First, the novel design of gled with not fitting in and frequently experienced depression
this study allowed the double empathy problem to be and anxiety. My special interest in Buddhism and meditation
explored in the context of EmpAcc as well as through an helped me not only to cultivate self-compassion, but also pro-
embodied felt sense of emotion. Second, unlike previous vided a profound clarity to navigate the complexities of my mind.
studies exploring the double empathy problem, our study As a meditation teacher, and through my research on autism,
used dynamic and more ecologically valid stimuli to mindfulness, empathy and compassion, I hope to bridge gaps in
understanding and empower autistic individuals to feel more
observe non-autistic participants’ ability to accurately
understood in a neurodiverse world. As a mother of two autistic
empathise with the emotions of autistic and non-autistic children, I am both saddened and motivated by the alarming sta-
narrators rather than just interpreting mental states or emo- tistics on suicidality within the autistic community, propelling
tion expression recognition. Finally, we used a relatively me to advocate for increased support and greater empathy. I
large sample meaning that the study was adequately pow- acknowledge that my positionality influences how my research is
ered. Despite the strengths of this study, we acknowledge conducted, the research questions, its outcomes and the interpre-
several limitations. Most critically, we only used partici- tation of the results.
pants from the general population instead of including
diagnosed autistic participants. This severely limited the Acknowledgements
inferences that could be drawn and only really tested one The authors would like to thank the narrators for their time and
side of the double empathy problem, a criticism made by participation in filming the video clips used in the EmpAcc task.
Chown et al. (2020). Furthermore, our study only involved
adults from a similar cultural background without intel- Declaration of conflicting interests
lectual disability. Future studies should include children The author(s) declared no potential conflicts of interest with
and adolescents and a range of cultural backgrounds. respect to the research, authorship, and/or publication of this
Moreover, although our study included male and female article.
narrators, all narrators were White, highly educated and
from stable socioeconomic backgrounds, it would be ben- Funding
eficial to include narrators from other ethnicities and back- The author(s) received no financial support for the research,
grounds in future studies. Finally, we did not examine authorship, and/or publication of this article.
individual differences, such as alexithymia, which are
known to strongly affect empathic abilities. ORCID iDs
In conclusion, this study provides compelling evidence
Rachael TS Cheang https://orcid.org/0000-0002-6418-7391
that people from the general population, including those
with High-AQ, Medium-AQ and Low-AQ, struggle to Maya Skjevling https://orcid.org/0009-0009-5657-2087
accurately empathise with autistic people compared to
non-autistic people. The non-autistic participants had sig- Supplemental material
nificantly lower EmpAcc scores overall when viewing Supplemental material for this article is available online.
autobiographical accounts of emotional events from autis-
tic narrators compared to non-autistic narrators. They were References
less able to track the emotions of autistic narrators than Baron-Cohen, S., Jolliffe, T., Mortimore, C., & Robertson, M.
non-autistic narrators unless the participant felt the emo- (1997). Another advanced test of theory of mind: Evidence
tion more intensely in the body as was the case with anger from very high functioning adults with autism or Asperger
Cheang et al. 11

syndrome. Journal of Child Psychology and Psychiatry, Capps, L., Kasari, C., Yirmiya, N., & Sigman, M. (1993).
and Allied Disciplines, 38(7), 813–822. https://doi.org/10. Parental perception of emotional expressiveness in chil-
1111/j.1469-7610.1997.tb01599.x dren with autism. Journal of Consulting and Clinical
Baron-Cohen, S., Leslie, A. M., & Frith, U. (1985). Does the Psychology, 61(3), 475–484. https://doi.org/10.1037//0022-
autistic child have a ‘theory of mind.’ Cognition, 21(1), 006x.61.3.475
37–46. https://doi.org/10.1016/0010-0277(85)90022-8 Chen, Y.-L., Senande, L. L., Thorsen, M., & Patten, K. (2021).
Baron-Cohen, S., & Wheelwright, S. (2004). The empa- Peer preferences and characteristics of same-group and
thy quotient: An investigation of adults with Asperger cross-group social interactions among autistic and non-
syndrome or high functioning autism, and normal sex autistic adolescents. Autism, 25(7), 1885–1900. https://doi.
differences. Journal of Autism and Developmental org/10.1177/13623613211005918
Disorders, 34(2), 163–175. https://doi.org/10.1023/B: Chown, N., Hughes, L., & Baker-Rogers, J. (2020). What
JADD.0000022607.19833.00 about the other side of double empathy? A response to
Baron-Cohen, S., Wheelwright, S. J., Skinner, R., Martin, J., Alkhaldi, Sheppard and Mitchell’s JADD article concern-
& Clubley, E. (2001). The Autism-Spectrum Quotient ing mind-reading difficulties in autism. Journal of Autism
(AQ): Evidence from Asperger syndrome/high-functioning and Developmental Disorders, 50(2), 683–684. https://doi.
autism, males and females, scientists and mathematicians. org/10.1007/s10803-019-04263-y
Journal of Autism and Developmental Disorders, 31(1), Cohen-Rottenberg, R. (2011). Unwarranted conclusions and
5–17. https://doi.org/10.1023/A:1005653411471 the potential for harm: My reply to Simon Baron-Cohen.
Begeer, S., Koot, H. M., Rieffe, C., Meerum Terwogt, M., Autism and Empathy. https://autismandempathyblog.word-
& Stegge, H. (2008). Emotional competence in children press.com/unwarranted-conclusions-and-the-potential-for-
with autism: Diagnostic criteria and empirical evidence. harm-my-reply-to-simon-baron-cohen/
Developmental Review, 28(3), 342–369. https://doi.org/10. Crompton, C. J., Ropar, D., Evans-Williams, C. V. M., Flynn,
1016/j.dr.2007.09.001 E. G., & Fletcher-Watson, S. (2020). Autistic peer-to-peer
Blair, R. J. R. (2005). Responding to the emotions of oth- information transfer is highly effective. Autism, 24(7),
ers: Dissociating forms of empathy through the study of 1704–1712. https://doi.org/10.1177/1362361320919286
typical and psychiatric populations. Consciousness and Cuff, B. M. P., Brown, S. J., Taylor, L., & Howat, D. J. (2016).
Cognition, 14(4), 698–718. https://doi.org/10.1016/j.con- Empathy: A review of the concept. Emotion Review, 8(2),
cog.2005.06.004 144–153. https://doi.org/10.1177/1754073914558466
Blair, R. J. R. (2008). Fine cuts of empathy and the amyg- Davis, M. H. (1983). Measuring individual differences in empa-
dala: Dissociable deficits in psychopathy and autism. The thy: Evidence for a multidimensional approach. Journal of
Quarterly Journal of Experimental Psychology, 61(1), 157– Personality and Social Psychology, 44(1), 113–126. https://
170. https://doi.org/10.1080/17470210701508855 doi.org/10.1037/0022-3514.44.1.113
Bolis, D., Lahnakoski, J. M., Seidel, D., Tamm, J., & Schilbach, Dvash, J., & Shamay-Tsoory, S. G. (2014). Theory of mind and
L. (2021). Interpersonal similarity of autistic traits pre- empathy as multidimensional constructs: Neurological
dicts friendship quality. Social Cognitive and Affective foundations. Topics in Language Disorders, 34(4), 282–
Neuroscience, 16(1–2), 222–231. https://doi.org/10.1093/ 295. https://doi.org/10.1097/TLD.0000000000000040
scan/nsaa147 Edey, R., Cook, J., Brewer, R., Johnson, M. H., Bird, G., & Press,
Bollen, C. (2023). A reflective guide on the meaning of empa- C. (2016). Interaction takes two: Typical adults exhibit
thy in autism research. Methods in Psychology, 8, Article mind-blindness towards those with autism spectrum dis-
100109. https://doi.org/10.1016/j.metip.2022.100109 order. Journal of Abnormal Psychology, 125(7), 879–885.
Brewer, R., Biotti, F., Catmur, C., Press, C., Happé, F. G. E., https://doi.org/10.1037/abn0000199
Cook, R., & Bird, G. (2016). Can neurotypical individu- Eisenberg, N., & Miller, P. A. (1987). The relation of empathy
als read autistic facial expressions? Atypical production of to prosocial and related behaviors. Psychological Bulletin,
emotional facial expressions in autism spectrum disorders. 101(1), 91–119. https://doi.org/10.1037/0033-2909.101.1.91
Autism Research, 9(2), 262–271. https://doi.org/10.1002/ Faso, D. J., Sasson, N. J., & Pinkham, A. E. (2015). Evaluating
aur.1508 posed and evoked facial expressions of emotion from
Brown, M. B., & Forsythe, A. B. (1974). Robust tests for the adults with autism spectrum disorder. Journal of Autism
equality of variances. Journal of the American Statistical and Developmental Disorders, 45(1), 75–89. https://doi.
Association, 69(346), 364–367. https://doi.org/10.1080/016 org/10.1007/s10803-014-2194-7
21459.1974.10482955 Fletcher-Watson, S., & Bird, G. (2020). Autism and empathy:
Camm-Crosbie, L., Bradley, L., Shaw, R., Baron-Cohen, What are the real links? Autism, 24(1), 3–6. https://doi.
S., & Cassidy, S. (2019). ‘People like me don’t get org/10.1177/1362361319883506
support’: Autistic adults’ experiences of support and Gernsbacher, M. A., Stevenson, J. L., & Dern, S. (2017).
treatment for mental health difficulties, self-injury and sui- Specificity, contexts, and reference groups matter when
cidality. Autism, 23(6), 1431–1441. https://doi.org/10.1177/ assessing autistic traits. PLOS ONE, 12(2), Article
1362361318816053 e0171931. https://doi.org/10.1371/journal.pone.0171931
Camus, L., Macmillan, K., Rajendran, G., & Stewart, M. (2022). Gernsbacher, M. A., & Yergeau, M. (2019). Empirical failures
‘I too, need to belong’: Autistic adults’ perspectives on of the claim that autistic people lack a theory of mind.
misunderstandings and well-being. PsyArXiv. https://doi. Archives of Scientific Psychology, 7(1), 102–118. https://
org/10.31234/osf.io/5mysh doi.org/10.1037/ARC0000067
12 Autism 00(0)

Grace, K., Remington, A., Lloyd-Evans, B., Davies, J., & autism spectrum: A meta-analytic review. Neuroscience
Crane, L. (2022). Loneliness in autistic adults: A sys- and Biobehavioral Reviews, 90, 146–163. https://doi.
tematic review. Autism, 26(8), 2117–2135. https://doi. org/10.1016/j.neubiorev.2018.04.009
org/10.1177/13623613221077721 Lombardo, M. V., Barnes, J. L., Wheelwright, S. J., & Baron-
Grossman, R. B., Edelson, L. R., & Tager-Flusberg, H. (2013). Cohen, S. (2007). Self-referential cognition and empa-
Emotional facial and vocal expressions during story retelling thy in autism. PLOS ONE, 2(9), Article e883. https://doi.
by children and adolescents with high-functioning autism. org/10.1371/journal.pone.0000883
Journal of Speech, Language, and Hearing Research, 56(3), Macdonald, H., Rutter, M., Howlin, P., Rios, P., Le Conteur, A.,
1035–1044. https://doi.org/10.1044/1092-4388(2012/12- Evered, C., & Folstein, S. (1989). Recognition and expres-
0067) sion of emotional cues by autistic and normal adults. Journal
Happé, F. G. E. (1994). An advanced test of theory of mind: of Child Psychology and Psychiatry, and Allied Disciplines,
Understanding of story characters’ thoughts and feelings 30(6), 865–877. https://doi.org/10.1111/j.1469-7610.1989.
by able autistic, mentally handicapped, and normal children tb00288.x
and adults. Journal of Autism and Developmental Disorders, Mackes, N. K., Golm, D., O’Daly, O. G., Sarkar, S., Sonuga-
24(2), 129–154. https://doi.org/10.1007/BF02172093 Barke, E. J. S., Fairchild, G., & Mehta, M. A. (2018).
Harrison, J. L., Brownlow, C. L., Ireland, M. J., & Piovesana, Tracking emotions in the brain – Revisiting the Empathic
A. M. (2022). Empathy measurement in autistic and Accuracy Task. NeuroImage, 178, 677–686. https://doi.
nonautistic adults: A COSMIN systematic literature org/10.1016/j.neuroimage.2018.05.080
review. Assessment, 29(2), 332–350. https://doi.org/10.11 Magnée, M. J. C. M., De Gelder, B., Van Engeland, H., & Kemner,
77/1073191120964564 C. (2007). Facial electromyographic responses to emotional
Heasman, B., & Gillespie, A. (2019). Neurodivergent intersub- information from faces and voices in individuals with per-
jectivity: Distinctive features of how autistic people create vasive developmental disorder. Journal of Child Psychology
shared understanding. Autism, 23(4), 910–921. https://doi. and Psychiatry, and Allied Disciplines, 48(11), 1122–1130.
org/10.1177/1362361318785172 https://doi.org/10.1111/j.1469-7610.2007.01779.x
Hedley, D., Uljarević, M., Foley, K.-R., Richdale, A., & Trollor, Martin-Key, N. A., Brown, T., & Fairchild, G. (2017). Empathic
J. (2018). Risk and protective factors underlying depres- accuracy in male adolescents with conduct disorder and
sion and suicidal ideation in autism Spectrum Disorder. higher versus lower levels of callous-unemotional traits.
Depression and Anxiety, 35(7), 648–657. https://doi. Journal of Abnormal Child Psychology, 45(7), 1385–1397.
org/10.1002/da.22759 https://doi.org/10.1007/s10802-016-0243-8
Hill, E. L., & Frith, U. (2003). Understanding autism: Insights Maxwell, S. E., & Delaney, H. D. (2004). Designing experiments
from mind and brain. Philosophical Transactions of the and analyzing data: A model comparison perspective (2nd
Royal Society of London. Series B, Biological Sciences, ed.). Lawrence Erlbaum Associates Publishers.
358(1430), 281–289. https://doi.org/10.1098/rstb.2002. Mazurek, M. O. (2014). Loneliness, friendship, and well-being
1209 in adults with autism spectrum disorders. Autism, 18(3),
Hurst, R. M., Mitchell, J. T., Kimbrel, N. A., Kwapil, T. K., & 223–232. https://doi.org/10.1177/1362361312474121
Nelson-Gray, R. O. (2007). Examination of the reliabil- Mazza, M., Pino, M. C., Mariano, M., Tempesta, D., Ferrara,
ity and factor structure of the Autism Spectrum Quotient M., De Berardis, D., Masedu, F., & Valenti, M. (2014).
(AQ) in a non-clinical sample. Personality and Individual Affective and cognitive empathy in adolescents with autism
Differences, 43(7), 1938–1949. https://doi.org/10.1016/j. spectrum disorder. Frontiers in Human Neuroscience, 8,
paid.2007.06.012 1–6. https://doi.org/10.3389/fnhum.2014.00791
Ickes, W., Stinson, L., Bissonnette, V., & Garcia, S. (1990). McDonald, P. W., & Prkachin, K. M. (1990). The expression
Naturalistic social cognition: Empathic accuracy in and perception of facial emotion in alexithymia: A pilot
mixed-sex dyads. Journal of Personality and Social study. Psychosomatic Medicine, 52(2), 199–210. https://
Psychology, 59(4), 730–742. https://doi.org/10.1037/0022- doi.org/10.1097/00006842-199003000-00007
3514.59.4.730 Mckenzie, K., Russell, A., Golm, D., & Fairchild, G. (2022).
Jones, A. P., Happé, F. G. E., Gilbert, F., Burnett, S., & Viding, Empathic accuracy and cognitive and affective empathy in
E. (2010). Feeling, caring, knowing: Different types of young adults with and without autism spectrum disorder.
empathy deficit in boys with psychopathic tendencies and Journal of Autism and Developmental Disorders, 52(5),
autism spectrum disorder. Journal of Child Psychology and 2004–2018. https://doi.org/10.1007/s10803-021-05093-7
Psychiatry, and Allied Disciplines, 51(11), 1188–1197. Milton, D., Gurbuz, E., & Lopez, B. (2022). The ‘double empa-
https://doi.org/10.1111/J.1469-7610.2010.02280.X thy problem’: Ten years on. Autism, 26(8), 1901–1903.
Kanner, L. (1943). Autistic disturbances of affective contact. The https://doi.org/10.1177/13623613221129123
Nervous Child, 2(3), 217–250. Milton, D., Heasman, B., & Sheppard, E. (2018). Double
Kinnaird, E., Stewart, C., & Tchanturia, K. (2019). Investigating empathy. In Volkmar F. R. (Eds), Encyclopedia of
alexithymia in autism: A systematic review and meta- autism spectrum disorders (pp. 1509–1517). Springer
analysis. European Psychiatry, 55, 80–89. https://doi. International Publishing. https://doi.org/10.1007/978-3-
org/10.1016/j.eurpsy.2018.09.004 319-91280-6_102273
Leppanen, J., Sedgewick, F., Treasure, J., & Tchanturia, K. Milton, D. E. M. (2012). On the ontological status of autism:
(2018). Differences in the theory of mind profiles of The ‘double empathy problem’. Disability & Society, 27(6),
patients with anorexia nervosa and individuals on the 883–887. https://doi.org/10.1080/09687599.2012.710008
Cheang et al. 13

Morrison, K. E., DeBrabander, K. M., Jones, D. R., Faso, D. J., population sample of 6,900 typical adult males and
Ackerman, R. A., & Sasson, N. J. (2020). Outcomes of real- females. Molecular Autism, 6(1), Article 2. https://doi.
world social interaction for autistic adults paired with autis- org/10.1186/2040-2392-6-2
tic compared to typically developing partners. Autism, 24(5), Sachs, M. E., Kaplan, J., & Habibi, A. (2019). Echoing the emo-
1067–1080. https://doi.org/10.1177/1362361319892701 tions of others: Empathy is related to how adults and chil-
Moseley, R. L., Turner-Cobb, J. M., Spahr, C. M., Shields, dren map emotion onto the body. Cognition and Emotion,
G. S., & Slavich, G. M. (2021). Lifetime and perceived 33(8), 1639–1654. https://doi.org/10.1080/02699931.2019.
stress, social support, loneliness, and health in autistic 1591938
adults. Health Psychology, 40(8), 556–568. https://doi. Santiesteban, I., Gibbard, C., Drucks, H., Clayton, N., Banissy,
org/10.1037/hea0001108 M. J., & Bird, G. (2021). Individuals with autism share oth-
Mul, C.-L., Stagg, S. D., Herbelin, B., & Aspell, J. E. (2018). The ers’ emotions: Evidence from the continuous affective rating
feeling of me feeling for you: Interoception, alexithymia and and empathic responses (CARER) task. Journal of Autism
empathy in autism. Journal of Autism and Developmental and Developmental Disorders, 51(2), 391–404. https://doi.
Disorders, 48(9), 2953–2967. https://doi.org/10.1007/ org/10.1007/s10803-020-04535-y
s10803-018-3564-3 Sassenrath, C. (2020). ‘Let me show you how nice I am’:
Nicolaidis, C., Milton, D., Sasson, N. J., Sheppard, E. L., & Impression management as bias in empathic responses.
Yergeau, M. (2019). An expert discussion on autism and Social Psychological and Personality Science, 11(6), 752–
empathy. Autism in Adulthood: Challenges and Management, 760. https://doi.org/10.1177/1948550619884566
1(1), 4–11. https://doi.org/10.1089/aut.2018.29000.cjn Sasson, N. J., & Bottema-Beutel, K. (2022). Studies of autistic
Nummenmaa, L., Glerean, E., Hari, R., & Hietanen, J. K. (2014). traits in the general population are not studies of autism.
Bodily maps of emotions. Proceedings of the National Autism, 26(4), 1007–1008. https://doi.org/10.1177/13623
Academy of Sciences of the United States of America, 613211058515
111(2), 646–651. https://doi.org/10.1073/pnas.1321664111 Sasson, N. J., Faso, D. J., Nugent, J., Lovell, S., Kennedy, D.
Raymaker, D. M., Teo, A. R., Steckler, N. A., Lentz, B., Scharer, P., & Grossman, R. B. (2017). Neurotypical peers are less
M., Delos Santos, A., Kapp, S. K., Hunter, M., Joyce, A., & willing to interact with those with autism based on thin
Nicolaidis, C. (2020). ‘Having all of your internal resources slice judgments. Scientific Reports, 7, 1–10. https://doi.
exhausted beyond measure and being left with no clean- org/10.1038/srep40700
up crew’: Defining autistic burnout. Autism in Adulthood: Sheppard, E., Pillai, D., Wong, G. T.-L., Ropar, D., & Mitchell,
Challenges and Management, 2(2), 132–143. https://doi. P. (2016). How easy is it to read the minds of people
org/10.1089/aut.2019.0079 with autism spectrum disorder? Journal of Autism and
Rogers, K., Dziobek, I., Hassenstab, J., Wolf, O. T., & Convit, Developmental Disorders, 46(4), 1247–1254. https://doi.
A. (2007). Who cares? Revisiting empathy in Asperger syn- org/10.1007/s10803-015-2662-8
drome. Journal of Autism and Developmental Disorders, Volker, M. A., Lopata, C., Smith, D. A., & Thomeer, M. L.
37(4), 709–715. https://doi.org/10.1007/s10803-006-0197-8 (2009). Facial encoding of children with high-functioning
Rum, Y., & Perry, A. (2020). Empathic accuracy in clinical pop- autism spectrum disorders. Focus on Autism and Other
ulations. Frontiers in Psychiatry, 11, Article 457. https:// Developmental Disabilities, 24(4), 195–204. https://doi.
doi.org/10.3389/fpsyt.2020.00457 org/10.1177/1088357609347325
Ruzich, E., Allison, C., Smith, P., Watson, P., Auyeung, B., Zaki, J., Bolger, N., & Ochsner, K. (2008). It takes two: The
Ring, H., & Baron-Cohen, S. (2015). Measuring autis- interpersonal nature of empathic accuracy. Psychological
tic traits in the general population: A systematic review Science, 19(4), 399–404. https://doi.org/10.1111/j.1467-
of the Autism-Spectrum Quotient (AQ) in a nonclinical 9280.2008.02099.x