Autism and Empathy: Are We Measuring the Wrong Thing?
Autism and Empathy: Are We Measuring the Wrong Thing?
The Society Journal | Research Insight | November 2026
What 226 studies and a new generation of research tell us about understanding, feeling and caring
Abstract
The idea that autistic people lack empathy has shaped autism research, diagnosis and public perception for decades.
Contemporary evidence presents a substantially more complicated picture.
A major 2025 systematic review and meta analysis synthesising 226 studies found a large average difference between autistic and non autistic participants on measures classified as cognitive empathy, but only a small difference in affective empathy. When analyses of affective empathy were restricted to higher quality studies, that difference was no longer statistically significant. Crucially, the apparent size of the “empathy deficit” changed substantially according to how empathy was measured.
A 2026 systematic review focused on children and adolescents reached a similar conclusion. Most of the included studies reported difficulties in cognitive empathy alongside relatively intact affective empathy. Meanwhile, newer experimental studies suggest that some autistic people may experience strong emotional reactions to others while finding it more difficult to identify, interpret or translate social and emotional signals into the responses expected by observers.
Research into the double empathy problem has also challenged the assumption that misunderstanding in autistic and non autistic interactions necessarily originates entirely within the autistic person. Yet evidence for this theory remains nuanced, and methodological reviews caution against claiming that all empathy differences are simply bidirectional misunderstandings.
The question is therefore becoming more precise.
It is no longer simply:
“Do autistic people have empathy?”
It is:
“Which processes are we calling empathy, how are we measuring them, and are difficulties in interpreting another person’s emotions the same thing as not caring about them?”
This Research Insight examines how modern evidence is reshaping one of autism’s most persistent assumptions.
Keywords: autism, empathy, cognitive empathy, affective empathy, emotion recognition, double empathy problem, neurodiversity, social cognition
Editorial approach
This article is a narrative research synthesis rather than a systematic review.
It draws primarily on a 2025 systematic review and meta analysis of 226 studies, a 2026 systematic review in children and adolescents, recent experimental research into empathic accuracy and emotion processing, and methodological work examining the double empathy problem.
Throughout this article, empathy is treated as a multidimensional construct rather than a single ability.
That distinction is essential.
The problem begins with the word “empathy”
To ask whether autistic people “have empathy” sounds like a straightforward scientific question.
It is not.
Researchers do not agree on one universally accepted definition of empathy.
The 2025 meta analysis by Cusson and colleagues noted that previous scholarship had identified 31 different definitions of empathy within autism research, alongside dozens of different methods used to measure it.
That alone should make simple conclusions difficult.
One study might define empathy as recognising another person’s emotion.
Another might measure the ability to infer what somebody is thinking.
Another might ask whether someone emotionally responds to another person’s distress.
Another might use a self report questionnaire asking how compassionate the participant believes themselves to be.
These are related abilities.
They are not identical.
Yet historically, results from very different tasks have often been collapsed into one broad conclusion:
autistic people have an empathy deficit.
The latest evidence suggests that this conclusion may be too imprecise to be scientifically useful.
Cognitive empathy and affective empathy are not the same thing
One of the most important distinctions in contemporary empathy research is between cognitive empathy and affective empathy.
Cognitive empathy generally refers to the ability to identify or understand another person's emotional or mental state.
It involves questions such as:
What is this person feeling?
Why might they be feeling it?
What are they thinking?
What do they intend?
Affective empathy concerns the emotional response generated in the observer.
Do I emotionally resonate with this person's distress?
Do I feel concern for them?
Does their emotional experience generate a response in me?
A person could theoretically have difficulty identifying exactly why another person is upset while nevertheless being deeply distressed by the fact that the person is suffering.
Conversely, someone might be highly skilled at identifying what another person feels without experiencing much emotional concern themselves.
Treating both abilities as a single trait can therefore obscure important differences.
What 226 studies actually found
In 2025, Cusson and colleagues published one of the most comprehensive examinations of autism and empathy to date.
Their systematic review and meta analysis included 226 studies.
The headline result was striking.
For cognitive empathy, autistic participants showed a large average difference relative to non autistic participants:
Hedges' g = -0.85.
For measures that treated empathy as one undivided construct, the average difference was even larger:
g = -1.70.
But for affective empathy, the average difference was much smaller:
g = -0.17.
When the researchers restricted the affective empathy analysis to higher quality studies, that difference was no longer statistically significant.
This does not show that every autistic person possesses intact affective empathy.
Autistic people are heterogeneous.
Some individuals may experience significant difficulties in both cognitive and affective components.
But the meta analysis challenges the much broader assertion that autism is characterised by a general inability to emotionally care about or respond to others.
The measurement itself changes the answer
Perhaps the most important finding from the 2025 meta analysis was methodological.
The apparent magnitude of autism related empathy differences changed depending on the measure researchers used.
This raises a fundamental question:
Are we measuring empathy, or are we sometimes measuring particular social cognitive abilities and calling them empathy?
Consider a widely used type of task in autism research.
A participant may be shown a photograph of someone's eyes and asked to select which mental or emotional state the person is experiencing.
Success requires multiple processes:
interpreting facial cues
understanding the vocabulary provided
choosing among predefined categories
making a rapid social inference
understanding the assumptions built into the task
A lower score may reflect difficulty with emotion recognition or social inference.
But does it necessarily establish an absence of emotional concern?
No.
The distinction matters because public language often moves directly from:
“This person has difficulty reading emotional cues”
to:
“This person lacks empathy.”
The science increasingly suggests that these statements should not be treated as equivalent.
Even within affective empathy, the picture is complicated
One of the most revealing examples comes from the Interpersonal Reactivity Index, a widely used multidimensional empathy questionnaire.
The 2025 meta analysis found different patterns depending on which affective empathy subscale was examined.
Autistic participants showed lower average empathic concern but higher personal distress than non autistic participants.
For personal distress, the effect was in the opposite direction:
g = 0.67.
That matters because a person who becomes highly distressed when witnessing another person's emotional pain cannot easily be described as emotionally indifferent.
Indeed, intense emotional responsiveness might itself complicate social responding.
A person may care strongly while finding the emotional situation overwhelming.
The observable response may then look very different from what an observer expects empathy to look like.
Children show a similar pattern
A 2026 systematic review by Lemos, Brandão and Diniz examined empathy specifically among autistic children and adolescents.
From an initial 1,081 records, only 15 studies met the review criteria. Most samples consisted predominantly of adolescent boys.
Ten of the 15 studies reported:
relatively intact affective empathy
alongside
differences or difficulties in cognitive empathy.
Five studies reported difficulties in both domains.
Again, the result is not:
all autistic children have intact emotional empathy.
It is that the evidence does not support treating empathy as one uniformly impaired capacity.
Different components appear to behave differently.
What if emotional response is actually stronger?
Some contemporary research goes further.
In October 2026, Wood and colleagues published a multimodal study examining empathy related processes among children aged 6 to 13 years.
The study included:
103 autistic children
and
40 non autistic children.
Participants watched videos showing young people experiencing strong emotions such as sadness and anger.
Researchers measured emotional responses, emotion recognition, perspective taking, physiological arousal and visual attention.
The results complicate the reduced empathy model considerably.
Autistic children showed greater affective intensity when viewing strongly emotional scenes, but they were less accurate on average at identifying the emotion being expressed.
Greater physiological and emotional intensity was associated with more frequent shifts of attention.
Those shifts were, in turn, associated with less accurate emotion recognition.
The researchers proposed an Emotion Regulation and Recognition Account.
Under this emerging model, some autistic people may experience strong emotional responses to others but have greater difficulty regulating that response while simultaneously interpreting complex emotional cues.
The important implication is clear.
An outwardly unexpected response does not necessarily mean an absence of emotional engagement.
It may reflect too much emotional activation rather than too little.
Feeling is not the same as interpreting
This distinction may help explain an apparent contradiction that has persisted in autism research.
Many autistic people report feeling other people's emotions intensely.
At the same time, laboratory studies frequently identify difficulties with:
emotion recognition
perspective taking
mental state inference
rapid interpretation of social signals
These findings can coexist.
A person might sense that someone is distressed without confidently understanding why.
They might feel another person's distress strongly but not know what response is socially expected.
They may recognise the emotion only after additional time or context.
Or they may express concern in a way the other person does not recognise as empathic.
The mistake occurs when difficulty translating social information is assumed to mean absence of caring.
Autistic people themselves describe a more complicated experience
Empathy research has historically relied heavily on experimental tasks, questionnaires or reports from relatives and clinicians.
Less research has simply asked autistic people how they understand their own empathic experience.
Campbell Templeton and colleagues addressed that gap in research published in 2026.
The study involved 100 participants and explored self perceived empathy within the broader context of self concept.
Autistic participants gave themselves lower average empathy ratings than non autistic participants.
Yet qualitative analysis showed that participants from both groups described themselves as possessing both cognitive and affective empathic capacities.
The differences were more nuanced than the simple presence or absence of empathy.
The study is important for another reason.
If autistic people grow up repeatedly being told, explicitly or implicitly, that autism is associated with deficient empathy, those narratives may themselves influence self perception.
That possibility requires further research.
But it illustrates why scientific constructs cannot be entirely separated from the cultural meanings surrounding them.
Then there is the double empathy problem
In 2012, autistic researcher Damian Milton proposed what became known as the double empathy problem.
The argument challenges a one directional model of social difficulty.
Traditional models frequently ask:
Why does the autistic person fail to understand the non autistic person?
The double empathy framework asks:
What happens when two people with substantially different ways of communicating and experiencing the world try to understand one another?
Misunderstanding might be reciprocal.
An autistic person may struggle to interpret a non autistic person's communication.
But a non autistic person may also struggle to interpret an autistic person's communication.
Social difficulty therefore may sometimes arise between people, rather than existing solely as a deficit inside one person.
What happens when autistic and non autistic people actually try to read each other?
A 2026 study in Molecular Autism directly examined this question using an empathic accuracy paradigm.
The study included:
141 autistic adults
and
94 non autistic adults.
Participants watched autobiographical videos from autistic and non autistic storytellers and attempted to track the storyteller's emotional state.
Researchers then compared participants' estimates with the storytellers' own reports.
There were no significant overall differences between autistic and non autistic participants in accuracy for continuous emotional valence or specific emotions.
There was a trend suggesting autistic participants might show relatively greater accuracy when evaluating autistic storytellers, but the interaction did not reach conventional statistical significance:
p = .059, d = 0.13.
Interestingly, non autistic participants reported higher self rated empathy and empathic interest, even though their objective empathic accuracy was not significantly better.
The researchers concluded that the results offered partial, not definitive, support for the double empathy account.
That distinction is important.
The double empathy problem is not a complete replacement theory
There is a temptation to replace one absolute statement:
autistic people have empathy deficits
with another:
autistic people have no empathy differences at all; non autistic people simply misunderstand them.
The evidence does not justify either extreme.
A 2026 critical methodological review by Fellowes examined 20 empirical papers frequently cited as evidence for the double empathy problem.
The review argued that many of these studies did not actually measure cognitive or affective empathy directly, making it difficult to use them as decisive evidence about empathy itself.
Fellowes has also argued that a neurodiversity compatible account of empathy can acknowledge two ideas simultaneously:
cross neurotype misunderstanding can be reciprocal
and
some autistic people may nevertheless experience genuine difficulties with particular forms of empathy.
That is arguably the more scientifically mature position.
Neurodiversity does not require pretending that no autistic person experiences difficulty.
It requires avoiding the assumption that every difference must be interpreted through a universal deficit model.
Who is judging whom?
A particularly revealing study published on 10 November 2026 examined how non autistic adults evaluated empathic responses produced by autistic and non autistic children.
So and Li gave 61 non autistic adults transcripts derived from responses by 82 autistic and non autistic children. The adults rated how well the children's responses demonstrated understanding, empathy and appropriate prosocial behaviour.
Autistic children's responses received lower empathy ratings.
But the study raises a fascinating complication.
Observers who had stronger autism knowledge grounded in a medical deficit framework showed an even larger difference in how they rated autistic versus non autistic responses.
The researchers concluded that autistic expressions of empathy may sometimes be misinterpreted by non autistic observers.
This does not prove that every apparent empathy difference is merely perception bias.
But it highlights something psychological research can easily overlook:
tests do not occur outside culture.
Observers bring expectations about what empathy is supposed to look like.
If autistic empathy is expressed differently, those expectations may influence how the response is interpreted.
Empathy can be present without looking conventional
Imagine someone tells a friend about a bereavement.
One person immediately says:
“I'm so sorry. That must be incredibly difficult.”
Another becomes quiet.
Another asks detailed practical questions.
Another shares a similar experience from their own life.
Another feels overwhelmed and temporarily withdraws.
Social conventions may interpret some of these responses as more empathic than others.
But behavioural appearance alone does not reveal the internal emotional state.
This does not mean every response is equally helpful to the grieving person.
Communication still matters.
But scientific measurement needs to distinguish:
what someone feels
what they understand
how they communicate it
and
how another person interprets that communication
Those are four different stages.
Traditional empathy research has sometimes collapsed them together.
Autism is heterogeneous here too
One risk of correcting the empathy stereotype is creating another stereotype:
autistic people actually have more empathy.
That claim is also too broad.
Some autistic people describe intense emotional responsiveness.
Others report difficulty understanding other people's emotions.
Some experience both simultaneously.
Some may genuinely experience reduced affective concern.
Others may experience high personal distress.
Autism encompasses substantial variation in language, cognition, sensory experience, emotional processing, co occurring conditions and social experience.
The correct scientific conclusion is therefore not that autistic people universally have intact or heightened empathy.
It is that a universal empathy deficit model does not adequately describe the evidence.
Alexithymia further complicates the picture
Another factor frequently discussed in empathy research is alexithymia, which refers to difficulty identifying or describing one's own emotions.
Alexithymia occurs in both autistic and non autistic populations but is reported at higher rates among autistic people.
This matters because recognising one's own emotional state may influence:
emotion regulation
interpretation of others' emotions
responses on empathy questionnaires
Some researchers have proposed that certain emotional processing differences attributed historically to autism may partly reflect co occurring alexithymia.
The 2025 meta analysis examined alexithymia as a potential moderator but did not find that it statistically explained the overall empathy differences across the included studies.
The broader point remains important.
Autism rarely operates as the only relevant psychological characteristic in an individual.
Empathy research that searches for one uniquely autistic explanation may overlook interacting traits and conditions.
The ethical consequence of getting this wrong
The empathy question is not merely academic.
The idea that autistic people lack empathy can influence how autistic individuals are perceived in:
schools
relationships
employment
healthcare
wider society
A person believed to be incapable of understanding or caring about others may be judged as:
cold
selfish
uncaring
rude
morally indifferent
That interpretation can persist even where the person is experiencing considerable emotional distress.
A scientific construct therefore becomes a social judgement.
This is why precision matters.
Difficulty recognising emotion is not synonymous with lack of compassion.
Difficulty predicting another person's perspective is not the same as indifference.
An unconventional response is not proof that no emotional response occurred.
What the evidence currently supports
WHAT WE KNOW
Empathy is multidimensional.
Cognitive empathy and affective empathy should not be treated as identical processes.
The largest recent meta analysis, involving 226 studies, found substantial average differences in cognitive empathy but only a small difference in affective empathy. The affective difference disappeared when analyses were restricted to higher quality studies.
Research in children and adolescents similarly suggests that cognitive empathy differences are more consistently reported than affective empathy differences.
WHAT THE RESEARCH SUGGESTS
Some autistic people may experience strong or even heightened emotional responses to other people's distress while simultaneously experiencing difficulty recognising, interpreting or responding to emotional cues.
Measurement choice substantially influences conclusions about autistic empathy.
Cross neurotype communication differences may also contribute to interpersonal misunderstanding.
Recent experimental evidence offers partial support for this double empathy framework, but not a complete explanation.
WHAT WE DO NOT YET KNOW
We do not have one universally accepted operational definition of empathy.
We do not know how well many laboratory measures reflect real world interpersonal understanding.
We cannot assume that empathy findings from verbally able autistic research samples generalise across the entire autistic population.
Research has not established that every empathy related difficulty observed in autism is either entirely intrinsic or entirely produced by cross neurotype misunderstanding.
What should be studied next?
The next phase of research needs to become considerably more ecologically valid.
Real interactions
Static facial expressions and multiple choice emotion tests reveal something about social cognition.
They do not fully reproduce human relationships.
Researchers need more studies involving genuine conversations, relationships and emotionally meaningful exchanges.
Separate the components
Studies should explicitly distinguish:
emotion recognition
perspective taking
affective resonance
personal distress
compassion
prosocial motivation
behavioural response
Calling all of these “empathy” creates conceptual confusion.
Include autistic perspectives
Autistic participants should not only complete empathy tests.
They should contribute to defining which forms of interpersonal understanding are meaningful and how empathy is experienced and expressed.
Examine both directions
If social understanding is relational, studies need to measure:
autistic understanding of non autistic people
non autistic understanding of autistic people
autistic understanding of autistic people
non autistic understanding of non autistic people
Greater representation
Research samples remain biased towards autistic people able to participate in conventional verbal and experimental studies.
Far less is known about empathy among autistic people with intellectual disabilities or different communication needs.
Conclusion: understanding emotion is not the same as caring
The statement that autistic people lack empathy appears increasingly difficult to defend as a single scientific proposition.
It collapses too many distinct processes into one label.
Contemporary evidence suggests that autistic people, on average, can experience greater difficulty with some forms of:
emotion recognition
social inference
perspective taking
But that does not automatically imply an absence of emotional resonance or concern.
Indeed, some recent evidence suggests precisely the opposite. Strong emotional responses may coexist with difficulties interpreting emotional cues or producing socially expected responses.
The double empathy problem further reminds us that interpersonal understanding occurs between people.
The observer is not neutral.
Communication has to be interpreted.
Autistic expression may be misread by people whose expectations were developed primarily through non autistic social norms.
None of this means that every autistic person possesses typical or heightened empathy.
Some experience substantial empathy related difficulties.
Autism is heterogeneous.
The more important conclusion is methodological.
Perhaps science has spent too long asking:
“How much empathy do autistic people have?”
when the better questions are:
“Which component of empathy are we measuring?”
“Whose emotional language is being treated as the standard?”
and
“Does difficulty understanding another person's emotion necessarily mean that we do not care about them?”
Once those questions are separated, the stereotype of the “unempathetic autistic person” becomes much harder to sustain.
The science has not discovered that empathy is absent.
It has discovered that empathy is more complicated than the tests designed to measure it.
The Society Research Position
WHAT WE KNOW
Autistic people show more consistent average differences in cognitive empathy and emotion recognition than in affective empathy.
WHAT THE RESEARCH SUGGESTS
Strong emotional responsiveness can coexist with difficulty interpreting social and emotional cues, and some interpersonal misunderstandings may be bidirectional across neurotypes.
WHAT WE DO NOT YET KNOW
No single empathy model adequately explains every autistic person, and current evidence does not justify either a universal empathy deficit model or the claim that all empathy differences are simply misunderstandings caused by non autistic observers.
WHAT SHOULD BE STUDIED NEXT
Research needs more naturalistic social interactions, clearer separation of empathy components, greater autistic involvement in study design, and broader representation of autistic people with different communication and support needs.
References
Cusson NM, Meilleur AJ, Bernhardt BC, Soulières I, Mottron L. A systematic review and meta analysis of empathy in autism: The influence of measures. Clinical Psychology Review. 2025;120:102623. doi: 10.1016/j.cpr.2025.102623.
Lemos D, Brandão T, Diniz E. Empathy in Children and Adolescents with Autism Spectrum Disorder: A Systematic Review. Review Journal of Autism and Developmental Disorders. Published 5 February 2026. doi: 10.1007/s40489-025-00536-8.
Rum Y, Feldman N, Genzer S, et al. Nuances of double empathy in autistic and non autistic people: examination using the empathic accuracy paradigm. Molecular Autism. 2026;17:34. doi: 10.1186/s13229-026-00723-2.
Fellowes S. Cognitive Empathy, Affective Empathy and Empirical Research on the Double Empathy Problem: A Critical Methodology Review. Journal of Autism and Developmental Disorders. 2026. doi: 10.1007/s10803-026-07315-2.
Fellowes S. Autism, Empathy and Neurodiversity. Psychiatric Clinics of North America. 2026;49(3):493–504. doi: 10.1016/j.psc.2025.12.004.
Campbell Templeton S, Branney P, Mitchell P. How do autistic people view their empathic capacity? British Journal of Developmental Psychology. 2026;44(1):99–117. doi: 10.1111/bjdp.70009.
Wood JJ, Orm S, Wood KS, et al. The emotion regulation and recognition account of empathy related processes in autism: a multimodal, multilevel case control study. Journal of Child Psychology and Psychiatry. Published online 10 October 2026. doi: 10.1111/jcpp.70249.
So WC, Li XH. Do Autistic Children Have Empathy? Perspectives From Non Autistic Perceivers and the Role of Their Understanding About Autism. Journal of Autism and Developmental Disorders. Published online 10 November 2026. doi: 10.1007/s10803-026-07594-9.
The Society Journal publishes research led analysis across neurodivergence, science, health, society, education and culture. Research Insights distinguish established evidence from emerging findings, methodological limitations and assumptions that the evidence increasingly asks us to reconsider.


