What Is the Double Empathy Problem? Autism, Communication & Mutual Understanding

The double empathy problem is the idea that communication difficulties between autistic and non-autistic people are often mutual.

Instead of assuming that misunderstanding happens because the autistic person lacks social understanding, the theory asks us to look at the interaction between two people whose ways of experiencing, interpreting and expressing the world may differ.

An autistic person can misread a non-autistic person.

A non-autistic person can also misread an autistic person.

The difficulty can therefore exist between people, rather than entirely inside one of them.

🤝 Different communication styles meet
🧠 Each person interprets the other through their own expectations
💬 Tone, timing, expression and implied meaning may not align
🔄 Both people can misunderstand what the other intended
⚖️ Yet historically the autistic person has usually been treated as the one with the deficit

Autistic researcher Damian Milton introduced the term double empathy problem in 2012 as a challenge to theories that located autistic social difficulties almost exclusively within autistic people. His argument was not that autistic people never experience social or communication difficulties. It was that mutual understanding becomes harder when people have substantially different experiences and ways of making sense of the world. (Milton, 2012)

Since then, researchers have begun testing parts of this idea experimentally.

The evidence has produced something more interesting than either extreme.

Autistic communication is clearly not simply socially deficient.

But neither has research established a universal rule that autistic people always communicate better with autistic people or that every autistic–non-autistic misunderstanding can be explained by neurotype mismatch.

The science points toward something more nuanced: communication ability depends partly on who is communicating with whom, what they are doing, what each person expects and how well their communication styles align.

🤝 What Is the Double Empathy Problem?

Traditional explanations of autism often framed social difficulty as something located inside the autistic person.

An autistic person might:

💬 miss implied meaning
👀 use eye contact differently
⏱️ respond with different conversational timing
🧠 interpret social information differently
📚 give more detail than expected
😐 display emotion in ways other people do not recognise easily

These differences were commonly interpreted as evidence that autistic people had impaired social understanding.

The double empathy problem adds another question:

How good are non-autistic people at understanding autistic people?

That question changes the scientific frame.

If communication difficulty exists entirely because autistic people lack social ability, then we would expect autistic people to communicate poorly regardless of who they interact with.

But if part of the difficulty comes from a mismatch between communication styles, then performance should sometimes change depending on the interaction partner.

That idea has become one of the most influential shifts in modern autism communication research. It is also an important part of the broader change described in What Is Autism?: autistic communication can involve genuine differences and difficulties without every misunderstanding being evidence of an individual social deficit.

🧠 Where Did the Double Empathy Problem Come From?

Damian Milton introduced the term in his 2012 paper On the Ontological Status of Autism: The ‘Double Empathy Problem’.

His argument was partly philosophical.

People understand one another partly because they share assumptions, experiences and ways of interpreting the world.

When those experiences become very different, mutual understanding can become harder.

🌍 different lived experiences
🧠 different expectations
💬 different communication norms
👀 different interpretations of behaviour
🤝 less shared intuitive understanding

Milton described this as a disjuncture in reciprocity between people with different ways of experiencing the world. (Milton, 2012)

The original idea was therefore broader than the slogan it sometimes becomes online.

It was not simply:

“Autistic people understand autistic people and non-autistic people don’t.”

It was closer to:

Mutual understanding becomes harder when people with substantially different perspectives attempt to interpret one another, and autism research has historically placed responsibility for that mismatch almost entirely on the autistic person.

❤️ Does Autism Mean Low Empathy?

No.

The claim that autistic people simply lack empathy collapses several different psychological processes into one word.

Empathy can involve:

🧠 recognising what another person may be thinking
❤️ emotionally responding to another person’s feelings
👀 detecting emotional signals
🧩 interpreting the context behind those signals
🤝 caring about what happens to another person
💬 knowing how to express that concern in a socially expected way

A person can be strong in one of these areas and struggle in another.

Autistic people can care deeply about others while finding facial expressions, indirect communication or rapidly changing social information difficult to interpret.

Someone may understand that another person is distressed but not know what response is expected.

Someone else may feel another person’s distress intensely but struggle to identify exactly what caused it.

The opposite misunderstanding happens too: a non-autistic person may see reduced eye contact, a flat facial expression or a delayed response and incorrectly infer indifference.

That is precisely where double empathy becomes relevant.

Expression and experience are not the same thing.

For a related distinction, Alexithymia in Neurodivergent Adults explains why difficulty identifying or expressing emotions should not be confused with an absence of emotion or care.

🪞 Empathy Is Not the Same as Theory of Mind

Older autism research often focused on theory of mind: the ability to infer another person’s thoughts, intentions, beliefs or perspective.

Some studies found autistic group differences on particular theory-of-mind tasks.

Those findings became influential enough that autism was sometimes described almost as a disorder of social understanding.

The problem is that laboratory tasks do not capture all of real social interaction.

Real communication is:

🔄 reciprocal
⏱️ time-sensitive
👥 affected by both people
🧠 dependent on shared knowledge
🌍 shaped by culture and context
💬 expressed through language, timing, gesture, expression and tone simultaneously

The double empathy framework therefore shifts the unit of analysis.

Instead of asking only:

“How accurately can the autistic person understand another person?”

it also asks:

“How accurately can the other person understand the autistic person?”

That turns social understanding from a one-person test into an interaction.

💬 Why Communication Styles Can Become Mismatched

Consider a simple conversation.

One person communicates indirectly:

“It’s getting quite late.”

They expect the other person to infer:

“I would like you to leave.”

The other person interprets the literal informational content:

“Yes, it is late.”

Neither response is random.

They are operating with different assumptions about how meaning should be transmitted.

Other mismatches can involve:

💬 explicit language versus implication
⏱️ longer versus shorter response pauses
👀 different use of eye contact
😐 different facial expressiveness
📚 different expectations about how much detail is relevant
🔄 different ways of changing topics
🤐 different comfort with conversational silence
🎯 different assumptions about what the purpose of the conversation is

These differences do not automatically make either communication style defective.

Difficulty emerges when each person assumes their own style is the neutral one.

Research on autistic conversation increasingly supports the idea that interaction patterns vary with the partner and context rather than having one universally correct social rhythm. A 2026 study, for example, found longer silent gaps in autistic-autistic conversations but argued against treating one fixed turn-taking speed as the universal target. (Wehrle et al., 2026)

🔬 What Does the Research Actually Find?

The double empathy problem began as a theoretical critique.

Since then, several experimental research lines have tested predictions compatible with it.

The most influential findings involve:

📞 information transfer between people of the same and different neurotypes
🤝 rapport during autistic and non-autistic interactions
👀 first impressions of autistic adults
❤️ accuracy in interpreting autistic emotional experiences
🧠 how observers evaluate autistic–autistic versus mixed interactions

Several findings challenge a purely one-sided deficit account.

But they do not all point in exactly the same direction.

That distinction matters.

A good scientific explanation of double empathy should include both the supportive studies and the results that complicate the theory.

📞 The Famous Information-Transfer Study

One of the most influential tests was published by Catherine Crompton and colleagues in 2020.

Participants passed a story from one person to another, similar to the game sometimes called Telephone.

There were three kinds of chains:

♾️ autistic → autistic
👤 non-autistic → non-autistic
🔄 autistic ↔ non-autistic mixed chains

The researchers found that information transfer was similarly effective in autistic and non-autistic same-neurotype chains.

Mixed-neurotype chains showed poorer information transfer.

This was important because a simple autistic-deficit account would predict that chains made entirely of autistic participants should perform worst.

They did not.

Autistic-to-autistic information transfer was highly effective. (Crompton et al., 2020)

🧠 autistic people could transmit information effectively
🤝 same-neurotype autistic communication was not inherently impaired
🔄 mismatch appeared to matter in the original study
⚖️ social effectiveness therefore could not be explained by autism alone

This study became one of the strongest pieces of evidence associated with the double empathy problem.

It is also discussed in more detail in The Science of Autistic Communication.

🧪 But the 2025 Replication Changed the Picture

Science becomes stronger when influential findings are tested again.

In 2025, Crompton and a large international team published a preregistered replication and extension involving 311 participants across three sites.

They expected again to find poorer information transfer in mixed-neurotype chains.

They did not.

Information-transfer accuracy did not significantly differ between autistic, non-autistic and mixed-neurotype chains. (Crompton et al., 2025)

That finding is extremely important.

It means we should not present the 2020 result as establishing a universal rule that mixed-neurotype communication is inherently less accurate.

But the replication also produced another important result:

autistic-only chains again transferred information just as effectively as non-autistic chains.

So one part of the original challenge to the deficit model survived very well.

🧠 autistic communication was not globally less effective
📞 autistic groups transferred information accurately
🔄 the predicted mixed-neurotype information deficit did not replicate
📚 the field therefore needs a more nuanced model than either “autistic deficit” or “neurotype mismatch explains everything”

That is exactly how science should develop.

An interesting theory generates predictions.

Research supports some of them.

Larger studies then refine the theory.

🤝 What About Rapport?

Communication is about more than transmitting facts.

People also experience rapport: how comfortable, connected or coordinated an interaction feels.

A 2020 study found that neurotype matching influenced self-reported rapport. Autistic participants reported greater rapport when interacting with other autistic people than when interacting with non-autistic people. (Crompton et al., 2020)

Other real-world interaction research has similarly found differences in social experience depending on partner neurotype.

However, newer work again shows that the pattern is not perfectly simple.

The 2025 large replication found that non-autistic chains reported higher rapport overall, while diagnostic disclosure improved rapport. Other recent studies have found combinations of neurotype matching, diagnostic awareness and individual characteristics influencing how successful interactions feel. (Crompton et al., 2025)

So:

🤝 neurotype matching can affect rapport
♾️ some autistic people report greater comfort with autistic peers
👤 individual differences still matter enormously
🧠 diagnosis awareness can change interpretations
⚖️ same-neurotype does not guarantee connection

This matches lived experience surprisingly well.

Many autistic adults describe interaction with other autistic people as easier because less translation is required. A systematic review of autistic adults’ experiences with other autistic people found recurring themes of increased understanding, belonging and communication ease, although autistic-autistic relationships were not universally effortless or positive. (Watts et al., 2025)

For the lived relationship side, see Autistic Friendships in Adulthood.

👀 First Impressions: Non-Autistic People Also Misread Autism

Another important research line asks how autistic people are perceived.

This matters because social interaction begins before either person has said very much.

Studies have found that non-autistic observers can form less favourable first impressions of autistic adults from brief exposure to their behaviour.

Importantly, these impressions change depending on what information observers have.

Research by Sasson and colleagues found that informing people that someone was autistic and increasing autism knowledge improved some first impressions. (Sasson & Morrison, 2019)

A 2025 systematic review of first-impression research found that negative impressions of autistic people appear across several studies, but also emphasised that effects vary by the trait being judged, the information available and the methodology used. (Wanigasekera et al., 2025)

👀 autistic behaviour can be interpreted negatively very quickly
🧠 observers may attach inaccurate meaning to unfamiliar behaviour
📋 knowing someone is autistic can change interpretation
🌍 social outcomes therefore depend partly on the perceiver, not only the autistic person

That finding is difficult to reconcile with a theory in which all autistic social difficulties originate solely inside autistic cognition.

❤️ Can Non-Autistic People Read Autistic Emotions Accurately?

This is perhaps the most direct empathy question.

If autistic people are uniquely impaired at understanding others, then non-autistic people should generally be good at understanding autistic people.

Research suggests that this assumption is too simple.

A study examining empathic accuracy found evidence consistent with double empathy: people did not necessarily understand emotional experiences equally well across neurotypes. (Cheang et al., 2024/2025)

Other research has found that non-autistic observers can recognise that autistic people are expressing emotion while still having difficulty correctly identifying the context or meaning behind that expression.

This creates a subtle but important distinction.

Someone can see:

“This person is reacting strongly.”

while misunderstanding:

“Why are they reacting strongly?”

That is an empathy problem too.

But historically, mistakes made by non-autistic observers have rarely been described as a non-autistic social-communication deficit.

Double empathy exposes that asymmetry.

♾️ Do Autistic People Communicate Better With Autistic People?

Sometimes.

But not automatically.

Research supports the possibility that shared neurotype can improve comfort, rapport or mutual understanding in some situations.

It does not establish a rule that:

autistic + autistic = perfect communication

or:

autistic + non-autistic = poor communication

Autistic people differ enormously.

Two autistic people can have:

🧠 very different processing speeds
💬 opposite communication preferences
🎧 conflicting sensory needs
📚 different levels of detail preference
❤️ different emotional styles
🔄 different tolerance for spontaneity
⏱️ different conversational rhythms

An autistic person may therefore feel immediately understood by one autistic person and completely misunderstood by another.

Likewise, an autistic and non-autistic person who know one another well can develop extremely effective shared communication.

This is one reason the newer research is useful.

It pushes us away from replacing one stereotype with another.

Neurotype matters.

But so do familiarity, personality, context, shared experience, communication preferences and relationship history.

⚖️ What the Double Empathy Problem Does Not Prove

The double empathy framework is sometimes stretched much further than the evidence supports.

It does not prove:

🚫 that autistic people have no social-communication difficulties
🚫 that every misunderstanding is caused equally by both people
🚫 that autistic people always understand one another
🚫 that non-autistic people are inherently poor communicators
🚫 that theory-of-mind differences do not exist
🚫 that social-skills support is never useful
🚫 that all disability is created by society
🚫 that neurotype matching guarantees successful relationships

Some autistic people have significant difficulty interpreting other people’s intentions, navigating reciprocal conversation or adapting communication to context.

Those difficulties are real.

Double empathy does not need to erase them.

Its strongest contribution is narrower and more defensible:

autistic social difficulty cannot always be understood accurately by studying the autistic person in isolation.

Interaction itself matters.

🧩 Criticisms and Open Scientific Questions

The double empathy problem has become extremely influential, but scientists continue to debate how broad the theory is and how well its different claims have been demonstrated.

A 2025 derivation-chain analysis examined how the concept developed through the literature and argued that parts of the double-empathy evidence base require more theoretical precision and stronger empirical testing. (Livingston et al., 2025)

Recent researchers have also defended the framework while acknowledging that it needs refinement rather than treating it as a finished theory. (Bollen et al., 2026)

Important unanswered questions include:

🔬 Which parts of social interaction are most affected by neurotype mismatch?
🧠 Is shared neurotype itself important, or are shared communication styles the real mechanism?
🤝 How much does familiarity reduce mismatch?
🌍 How do culture, gender and language affect double empathy?
🗣️ Does the pattern generalise to autistic people with different communication and support needs?
📏 Which measures best capture genuine mutual understanding?
⏳ How do relationships change once people learn one another’s communication style?

This is not a weakness of science.

It is what an active scientific theory looks like.

🎭 Double Empathy and Autistic Masking

The double empathy problem also helps explain why masking can become so common.

If autistic communication is repeatedly interpreted through non-autistic expectations, autistic people may learn to translate themselves.

They may:

👀 force culturally expected eye contact
💬 add social cushioning to direct communication
😄 produce expected facial expressions
🤐 suppress natural responses
🧠 rehearse interactions
🎧 hide discomfort
🎭 imitate communication styles that other people understand more easily

This can reduce immediate misunderstanding.

But it also places most of the adaptation burden on one side of the interaction.

That is the link between double empathy and Autistic Masking.

A genuinely reciprocal model asks something different:

Why must the autistic person always become more understandable to the non-autistic person?

Could the non-autistic person also learn how autistic communication works?

That does not remove the value of flexible communication.

It makes flexibility mutual.

❤️ Double Empathy in Relationships

Double empathy can be particularly useful in close relationships because misunderstandings often become moral interpretations.

Imagine an autistic partner becoming quiet during conflict.

Their partner may interpret:

“You don’t care enough to respond.”

The autistic person may internally experience:

“I care so much that my processing has overloaded and I cannot find language.”

The observable behaviour is the same:

silence.

The meaning is completely different.

That does not automatically make either person’s feelings wrong.

One person experiences disconnection.

The other experiences loss of processing capacity.

The problem becomes much easier to address when both experiences are visible.

🤝 ask rather than assume intent
💬 make indirect expectations explicit
⏸️ allow processing time
📝 use written communication when useful
🧠 distinguish inability from unwillingness
❤️ explain emotional meaning rather than expecting it to be inferred

This is particularly relevant to Autistic Shutdowns in Adults and Stonewalling vs Autistic Shutdown in Relationships, where outwardly similar behaviour can arise from very different internal mechanisms.

💼 Double Empathy at Work

The same mismatch can appear in workplaces.

An autistic employee may communicate directly.

A manager interprets them as rude.

The employee believes they were simply being clear.

A manager gives an indirect instruction:

“It might be good to look at this again.”

They intend:

“This needs to be revised.”

The autistic employee interprets it as an optional suggestion.

The work is not changed.

Both people then believe the other communicated badly.

💬 implied instructions become explicit
📋 important information can be written down
🧠 managers explain expectations rather than relying on social inference
🤝 employees can explain preferred communication styles
⚖️ directness is not automatically interpreted as hostility

Research applying double-empathy ideas to employment suggests that workplace misunderstandings may reflect interpretation differences on both sides rather than an autistic social deficit alone. (Szechy et al., 2024)

🩺 Double Empathy in Healthcare and Therapy

Healthcare creates a particularly important version of the problem.

A clinician may expect distress to look like:

😢 visible emotion
👀 expected facial expressions
🗣️ spontaneous description of feelings
🤝 conventional rapport
⚡ quick responses to open-ended questions

An autistic patient may instead:

😐 describe severe distress with a flat expression
🧠 give highly factual information
⏳ need extra processing time
🫀 struggle to identify bodily or emotional states
👀 avoid eye contact in order to think
📝 communicate much more accurately in writing

If the clinician treats their own communication style as the neutral baseline, they can underestimate the patient’s distress or misunderstand their intentions.

This is one reason Neuroaffirming Care emphasises communication flexibility rather than demanding one normal form of interaction.

Research increasingly connects double-empathy thinking with neuroaffirmative support: intervention should not focus only on changing autistic behaviour, but can also improve environments, communication partners and mutual understanding. (Brosnan, 2025)

🌱 What Better Communication Actually Looks Like

The practical lesson of double empathy is not:

“Everyone should just accept misunderstanding.”

Nor is it:

“Autistic people never need communication support.”

It is that successful communication is a shared task.

Both people can become more understandable to the other.

💬 Say what you mean when precision matters
❓ Ask instead of assuming intention
⏳ Allow different processing speeds
📝 Use written communication when it improves accuracy
👀 Do not use eye contact as a test of attention or honesty
❤️ Explain emotional meaning when it may not be obvious
🤝 Learn the other person’s communication patterns over time
🔄 Repair misunderstandings without automatically assigning blame

This approach also explains why some long-term autistic–non-autistic relationships communicate extremely well.

They have developed a shared language.

The relevant distinction may eventually turn out to be less:

same neurotype versus different neurotype

and more:

shared versus unshared communication expectations.

That is an important question for future research.

🧭 A Better Way to Think About Autistic Communication

The most useful change produced by double empathy is not replacing:

“Autistic people are bad at social communication.”

with:

“Autistic people are actually perfect communicators.”

Both are simplistic.

A stronger conclusion is:

Communication is relational.

Autistic people bring particular communication styles, processing characteristics, strengths and difficulties into an interaction.

Non-autistic people bring theirs.

Sometimes those styles align.

Sometimes they do not.

And when communication breaks down, science should examine both people and the interaction between them before deciding where the problem resides.

That is a much larger change than it sounds.

For decades, autistic people were largely studied as the object of social misunderstanding.

The double empathy problem asks science to study the misunderstanding itself.

📌 Key Takeaways

🤝 The double empathy problem proposes that autistic–non-autistic misunderstandings can be reciprocal rather than originating only in the autistic person

🧠 Damian Milton introduced the concept in 2012 as a critique of one-sided deficit accounts of autistic social interaction. (Milton, 2012)

❤️ Autism does not simply mean “lacking empathy”; empathy includes several different cognitive, emotional and interpersonal processes

📞 A landmark 2020 study found highly effective autistic-to-autistic information transfer and poorer performance in mixed-neurotype chains. (Crompton et al., 2020)

🧪 A much larger preregistered 2025 replication again found effective autistic information transfer but did not replicate the poorer mixed-neurotype information-transfer effect. (Crompton et al., 2025)

🤝 Same-neurotype interaction can sometimes improve rapport and comfort, but the pattern is not universal

👀 Research shows that non-autistic observers can also misinterpret autistic behaviour and form less favourable first impressions

⚖️ Double empathy does not mean autistic people never have social-communication difficulties

🧩 The strongest conclusion is that communication ability is partly interaction-dependent

🎭 Understanding mutual mismatch can reduce pressure for autistic people to do all of the adapting through masking

🌱 Better communication involves mutual adjustment rather than one person continually translating themselves into the other’s norms

🔗 Read Next

🧠 What Is Autism? Brain Differences, the Spectrum and Lived Experience

🔬 The Science of Autistic Communication

🤝 Autistic Friendships in Adulthood

🎭 Autistic Masking

🧊 Autistic Shutdowns in Adults

❤️ Stonewalling vs Autistic Shutdown in Relationships

🧠 Alexithymia in Neurodivergent Adults

🌱 What Is Neuroaffirming Care?

♾️ Autism Learning Hub

📚 External Sources & Research

📚 Milton (2012) — On the Ontological Status of Autism: The ‘Double Empathy Problem’

📚 Milton, Gurbuz & López (2022) — The ‘Double Empathy Problem’: Ten Years On

📚 Crompton et al. (2020) — Autistic Peer-to-Peer Information Transfer Is Highly Effective

📚 Crompton et al. (2020) — Neurotype-Matching, but Not Being Autistic, Influences Self and Observer Ratings of Interpersonal Rapport

📚 Rifai et al. (2022) — Investigating Markers of Rapport in Autistic and Nonautistic Interactions

📚 Jones et al. (2024) — Non-autistic Observers Both Detect and Demonstrate the Double Empathy Problem

📚 Cheang et al. (2024/2025) — Do You Feel Me? Autism, Empathic Accuracy and the Double Empathy Problem

📚 Crompton et al. (2025) — Information Transfer Within and Between Autistic and Non-autistic People

📚 Wanigasekera et al. (2025) — First Impressions Towards Autistic People: A Systematic Review

📚 Livingston et al. (2025) — The Double Empathy Problem: A Derivation Chain Analysis and Cautionary Note

📚 Watts et al. (2025) — ‘A Certain Magic’: Autistic Adults’ Experiences of Interacting With Other Autistic People

📚 Brosnan (2025) — Neuro-affirmative Support for Autism, the Double Empathy Problem and Monotropism

📚 Wehrle et al. (2026) — Consistently Longer Silent Gaps in Autistic Speaker Pairs Across Conversational Contexts