Somewhere in the 1990s, a researcher named Simon Baron-Cohen published a theory that autistic people lacked a "theory of mind." The idea spread fast. It became the explanation for everything — why you struggled socially, why relationships felt impossible, why you always seemed to get it wrong.
The problem is, the theory was built on studies comparing autistic people with non-autistic people in non-autistic contexts. It was measuring one kind of mind against the standard of another, and calling the difference a deficit.
What is the double empathy problem?
The double empathy problem is a theory proposed by autistic researcher Dr Damian Milton in 2012. It argues that the social difficulties observed between autistic and non-autistic people are not caused by a deficit in the autistic person, but by a mutual mismatch in communication styles, social expectations, and ways of relating. Empathy requires shared context and similar dispositions to function well — and when two people have very different neurotypes, the breakdown runs in both directions. Neither party is deficient. Both are trying to understand someone whose experience of the world is substantially different from their own.1
What the research shows
- In a controlled study, autistic adults transferred information just as effectively as non-autistic adults — but detail was lost significantly in mixed autistic–non-autistic chains. The communication breakdown is between neurotypes, not within the autistic group.2
- A qualitative study found that autistic adults consistently described interaction with other autistic people as more comfortable, less exhausting, and characterised by a kind of understanding that cross-neurotype interaction rarely produced.3
- Non-autistic people form negative first impressions of autistic adults within seconds of brief video exposure — before any conversation — and are less willing to pursue social interaction. The barrier is not coming from one side.4
- Ten years on, a 2022 review confirmed the double empathy problem has moved from conceptual argument to empirically supported framework — shifting how the field understands autistic social difficulties from deficit to difference.5
Where the “no empathy” myth came from
The deficit framing of autistic social behaviour has deep roots. In 1985, Simon Baron-Cohen and colleagues proposed that autistic children showed a specific impairment in “theory of mind” — the ability to understand that other people have beliefs, intentions, and perspectives different from their own. The famous test was the Sally-Anne task: a marble is moved while one doll is absent; can the child correctly predict that the returning doll will look for it in the wrong place? Most autistic children failed the task. Baron-Cohen interpreted this as evidence of “mind-blindness” — a selective deficit in understanding other minds.
The implication cascaded outward quickly. If autistic people could not reliably understand other people’s mental states, then empathy — which requires modelling another person’s inner experience — would be structurally compromised. The word “deficit” became standard. It entered clinical training, diagnostic language, public understanding, and the minds of autistic people themselves — many of whom were told, directly or indirectly, that they were constitutionally limited in their capacity to connect with others.
Even Baron-Cohen has since walked back the original framing. In later years he acknowledged the mind-blindness theory was “too narrow” and “missed a lot, particularly the role of emotions.” The theory had something to it — autistic and non-autistic people do process and communicate social information differently — but the conclusion that this constituted a deficit was a category error. It assumed neurotypical social norms were the standard against which all communication should be measured. Anything that deviated from those norms was labelled failure. The double empathy problem dismantles exactly that assumption.
Two kinds of empathy — and why the distinction matters
One reason the “autistic people lack empathy” claim persisted so long is that empathy is not a single thing. Researchers distinguish between at least two forms: cognitive empathy (the intellectual ability to understand another person’s perspective or mental state) and affective empathy (the emotional experience of feeling what someone else feels, sometimes called emotional contagion).
The theory of mind research largely measured cognitive empathy — the ability to predict beliefs and intentions in structured tasks. Autistic people often score lower on these measures. But affective empathy is a different matter. Research consistently shows that many autistic people have intact or even heightened affective empathy. The feeling is there. The difficulty is often in the translation layer: knowing what to do with it, how to express it in ways others recognise, how to communicate care in forms that match the other person’s expectations.
For a significant proportion of autistic people, the experience is not diminished empathy but the opposite: too much. Emotional contagion — picking up and internalising others’ feelings without a filter — is a common autistic experience. The distress of someone nearby becomes your distress. A crowded room of stressed people is overwhelming not just sensorially but emotionally. A news headline about suffering can be physically destabilising. Hume and Burgess (2021) documented this directly: many autistic people self-report hyper-empathy rather than reduced empathy, and some of what looks like emotional withdrawal is actually a shutdown response to being overwhelmed by others’ feelings.6
The “cold and unfeeling” presentation that gets read as a lack of empathy is, in many cases, the aftermath of feeling too much. Shutdown as protection, not indifference. The absence of visible emotional response, not the absence of emotion.
What the double empathy problem actually argues
Milton’s argument, first set out in a 2012 paper in Disability & Society, is fundamentally relational. Empathy does not operate in a vacuum. It depends on shared context, shared reference points, shared ways of expressing and reading emotional states. When two people have sufficiently similar dispositions and life experiences, empathy flows relatively naturally. When their experiences of the world are substantially different, the work of mutual understanding becomes much harder — for both of them.1
The critical shift is the word “both.” The old model located the problem entirely within the autistic person. The double empathy problem locates it in the interaction — in the space between two people trying to understand each other across a significant neurological difference. Non-autistic people are not neutral observers standing at the correct social frequency. They are also trying to understand someone whose way of experiencing and communicating is unfamiliar to them. They are also failing — just in ways the research had not previously been designed to notice.
This is not a claim that autistic people have no social difficulties. Many do, and those difficulties are real. What the double empathy problem reframes is not the existence of the difficulty but its cause. The struggles are not evidence of a broken or incomplete social brain. They are the predictable outcome of two different communication styles meeting without a shared framework for translation — and without an environment that distributes responsibility for that translation fairly.
The evidence: what happens when autistic people communicate with each other
Since 2012, a growing body of experimental and qualitative evidence has supported Milton’s theory directly. The most controlled evidence comes from two studies by Crompton and colleagues at the University of Edinburgh, both published in 2020.
The first used a “diffusion chain” method — structurally similar to the children’s game of telephone — in which participants received a story and passed it along a chain of eight people. Three types of chains were run: all-autistic, all-non-autistic, and mixed. Autistic chains retained information just as effectively as non-autistic chains. The breakdown happened specifically in the mixed groups, where detail was lost at a significantly steeper rate. The conclusion was direct: autistic social differences are better understood as cross-neurotype interaction challenges rather than autistic deficits. The difficulty was not inside the autistic participants. It was between neurotypes.2
The second study, qualitative, asked autistic adults to describe their relationships with both autistic and non-autistic friends and family. The contrast was striking. Cross-neurotype friendships were described as valuable but consistently more effortful — “neurotypical friends make me tired, they don’t understand me. Even if it’s good it’s exhausting.” Within-neurotype connection was described differently: easier to follow, less draining, more comfortable. Communication styles matched. The effort of translation was absent. One participant articulated what many others described: “I never realised everybody felt as happy as I do when I am around autistic people.”3
These two studies together capture both dimensions of the double empathy problem: the quantitative demonstration that autistic communication works well peer-to-peer, and the lived experience of what it actually feels like when that friction is removed.
“I spent years thinking I was the problem in every failed social interaction. I wasn’t picking up signals. I wasn’t reading the room. I wasn’t warm enough. It took diagnosis and then finding this research to understand that the people I struggled to connect with were also struggling to connect with me. We were both lost. I just got all the blame.”
— Autistic adult, diagnosed autistic 2019.
The other side of the mismatch: non-autistic bias
A 2017 study by Sasson and colleagues at the University of Texas Dallas added a dimension to the picture that rarely makes it into public discussion. Non-autistic participants were shown brief video clips — “thin slices” of just a few seconds — of autistic and non-autistic adults in social interaction, and asked to rate their first impressions. Non-autistic participants consistently rated the autistic adults less favourably: as less likeable, more awkward, and less desirable as social partners. These judgments formed within seconds and did not change with increased exposure.4
The autistic adults were rated comparably on trustworthiness and intelligence. The negative impressions were specifically social. Something about autistic communication styles — not content, not intent, not character — was triggering immediate social rejection from non-autistic observers before any interaction had actually occurred.
This is the other half of the double empathy problem made visible. The difficulty in cross-neurotype interaction is not only autistic people struggling to navigate non-autistic social norms. Non-autistic people are also misreading autistic communication — rapidly, automatically, and with social consequences that fall almost entirely on the autistic person. The gap runs in both directions. The cost does not.
If the deficit story shaped how you understood yourself
The Unmasking Years works through exactly this: dismantling what you were told about yourself and building a more accurate account from lived experience. If you were diagnosed late and spent years believing the empathy myth, that work of reframing starts here. Read more about the book →
The masking loop: how being misread creates the presentation that gets misread
There is a mechanism here worth naming carefully, because it closes a loop that is hard to see from the outside.
When autistic communication is consistently misread as cold, awkward, or disinterested — by teachers, parents, peers, clinicians — the pressure to adapt is immediate and intense. Autistic people learn, often very young and without anyone naming the process, to suppress their natural communication style and replace it with something more legible to a non-autistic audience. This is masking. It works, to a point. It reduces friction. It allows passage through environments that would otherwise be actively hostile.
But masking has costs that compound over time. It requires continuous cognitive load. It creates a persistent gap between internal experience and outward presentation. It erodes identity, because when you perform a version of yourself for long enough, it becomes genuinely unclear what the unperformed version is. And it drives the burnout that is a defining experience of late-diagnosed autistic adults: the collapse that comes when the mask becomes too heavy to hold.
The masking loop is this: autistic communication gets misread, which produces social pressure to conform, which produces masking, which produces a presentation that looks less autistic and therefore “proves” the deficit was the problem in the first place. The non-autistic social environment creates the presentation it then uses to justify the demand for it. The double empathy framework makes this loop visible. Without it, autistic people are left trying to fix a problem that was never entirely theirs.
What this means in practice
The double empathy problem is not only a research reframe. It has direct implications for every environment in which autistic people interact with non-autistic people — which is most of them.
In therapy: Most mainstream social skills interventions for autistic people are built on the deficit model. They aim to teach autistic people to communicate in more neurotypical ways: more eye contact, smoother conversation turn-taking, more legible emotional expression. The double empathy problem challenges the premise. If the difficulty is a cross-neurotype mismatch rather than an autistic deficit, then interventions designed to make autistic people conform to one side of that mismatch are incomplete at best and harmful at worst. A more honest approach asks how the therapeutic environment itself accommodates autistic communication, rather than how the autistic person can be modified to fit a neurotypical standard.
In relationships: Mixed-neurotype relationships — friendships, partnerships, family — carry the cross-neurotype mismatch into the most personal contexts. Both people are genuinely trying. Both are finding parts of the interaction difficult. The double empathy problem does not resolve that difficulty, but it does redistribute responsibility for it. When an autistic person’s communication style is read as cold or withholding, the accurate question is not “how can the autistic person communicate more warmly” but “how can both people learn each other’s communication style more accurately.”
In workplaces: Misinterpretations of autistic communication in professional settings — directness read as rudeness, silence read as disengagement, pattern-focus read as inflexibility — produce exclusion that looks like performance failure. The double empathy framework reframes this as an accessibility issue rather than an individual deficit. Clear communication, explicit feedback structures, and reduced dependence on unspoken social inference are not accommodations for broken employees. They are what workplaces should offer to all cross-neurotype teams.
In diagnosis: The persistent use of deficit language in diagnostic criteria means that autistic people who do not match the “socially uninterested” stereotype — who are socially motivated but cross-neurotype communication is hard — may be missed or dismissed. Hume and Burgess (2021) noted explicitly that highly empathetic autistic people may be prevented from accessing diagnosis because the empathy myth is baked into the assessment tools themselves.6 If you were told you “don’t seem autistic” because you obviously care about people, this is why.
Why this matters for late-diagnosed autistic adults specifically
For autistic people who were diagnosed late — who spent decades trying to make sense of social difficulties without an accurate framework — the empathy myth did specific damage. It was not an abstract research position. It was a story absorbed directly, in therapy rooms, in school reports, in relationships, in the internal explanations constructed for why interactions kept going wrong.
“You don’t read people well.” “You can come across as cold.” “You don’t seem to notice how others feel.” These were delivered as character assessments. And without a framework that named the other side of that mismatch — the non-autistic person who also wasn’t reading you well, who was also forming negative impressions they couldn’t justify, who was also failing to understand — the only available conclusion was that the problem was yours.
Understanding the double empathy problem is part of reframing your past through an autistic lens. The social difficulties were real. The explanation that made you the sole source of them was not. The interactions that felt like failure were often genuinely difficult — not because you were broken, but because you were trying to bridge a significant neurological difference without vocabulary, framework, or support. And because the environment you were in consistently told you which side of that gap was responsible for fixing it.
That distinction — between “I am broken” and “I was unsupported in navigating a bidirectional mismatch” — is not a small reframe. It changes the story of your life. It changes what you owe the story. And it changes what you look for now.
What the double empathy problem does not mean
Precision matters here, because the theory is sometimes misread in ways that undermine it. The double empathy problem does not mean that all autistic social difficulties are caused by non-autistic failure. It does not mean autistic people have no room to develop social skills, or that cross-neurotype relationships are impossible. It does not absolve autistic people of any responsibility for how they communicate.
What it means is that the framing of autism as a one-way deficit — located entirely in the autistic person’s brain — is empirically wrong. The difficulty is mutual, bidirectional, and contextual. And interventions built on the false premise will predictably fail to address the actual problem, because they are solving for the wrong thing.
It also means that autistic–autistic connection is not a consolation prize. The ease of within-neurotype communication is documented and legitimate. The communities, friendships, and relationships that autistic people build with each other are not lesser for being within-neurotype. They are often the first spaces where autistic people experience connection without the translation tax — where being understood does not require effort. That matters. It is worth pursuing, not apologising for.
Where the field stands now
The ten-year review by Milton, Gurbuz and López (2022) surveyed the accumulation of evidence since the original 2012 paper. Their conclusion: the double empathy problem has moved from conceptual argument to empirically supported framework, and the dominant deficit-based account of autistic social difficulties has been substantially challenged — though not yet fully replaced — in mainstream clinical and research settings.5
The language of “social communication disorder” and “theory of mind deficits” still appears in diagnostic criteria, clinical training, and public discussion. The double empathy problem has not fully displaced it. But it has given autistic people, researchers, and increasingly clinicians a more accurate framework for what is actually happening when autistic and non-autistic people interact.
For late-diagnosed autistic adults, that accuracy matters beyond academic interest. It changes the story you were told about yourself. It changes what you owe that story — and what you don’t.
For more on how the masking that follows from being told your communication is a deficit builds toward burnout, the articles on unmasking after late diagnosis and recovering from autistic burnout go deeper. If difficulty identifying or naming your own emotional responses is part of your experience — particularly if you have been told you lack empathy but feel like you actually feel too much — the article on alexithymia and autism is directly relevant. And for the broader work of reframing your history through an autistic lens, this article covers that process in detail.
Key takeaways
- The double empathy problem, proposed by autistic researcher Damian Milton in 2012, argues that autistic social difficulties arise from a mutual mismatch between neurotypes — not from a deficit in autistic people alone.
- Empathy has two components: cognitive empathy (perspective-taking) and affective empathy (feeling what others feel). Research shows many autistic people have intact or heightened affective empathy — emotional overload is more common than emotional absence.
- Autistic people communicate just as effectively with other autistic people as non-autistic people do with each other. Breakdowns occur specifically in mixed-neurotype interactions.
- Non-autistic people form negative first impressions of autistic adults within seconds — before any conversation — and are less willing to pursue interaction. The social barrier is bidirectional; the cost is not.
- Masking is in large part a response to being consistently misread — creating a loop where autistic people suppress natural communication to survive environments that were designed for a different neurotype.
- The double empathy problem has direct implications for therapy, relationships, workplaces, and diagnosis. Interventions built on the deficit premise systematically fail to address the actual problem.
- For late-diagnosed autistic adults, this reframe matters personally: the social difficulties were real. The explanation that placed all responsibility within the autistic person was not accurate.
Further reading on HeyASD
- Reframing your past: understanding childhood through an autistic lens
- Unmasking after late diagnosis: why it’s harder than it sounds
- Late autism diagnosis: a guide for autistic adults
- Alexithymia and autism: when emotions are hard to name
- How to recover from autistic burnout
- Autism and depression: misdiagnosis, burnout, and healing
What is the double empathy problem in simple terms?
The double empathy problem says that when autistic and non-autistic people struggle to understand each other, the difficulty is coming from both sides — not just the autistic person. Empathy works better between people who share similar ways of experiencing and communicating. When two people have very different neurotypes, the gap in mutual understanding is real for both of them. The traditional view only counted the autistic person’s side of that gap as a problem. The double empathy problem counts both sides.
Do autistic people actually lack empathy?
No. The claim that autistic people lack empathy is not supported by current evidence. Many autistic people experience intense affective empathy — the emotional experience of feeling what others feel — and some report hyper-empathy, where others’ emotions become overwhelming. What autistic people sometimes struggle with is cognitive empathy in cross-neurotype contexts: predicting what a non-autistic person is thinking or feeling based on subtle cues. The “no empathy” framing was a misreading of a cross-neurotype communication mismatch, attributed entirely to one side.
Who created the double empathy problem theory?
The double empathy problem was named and formalised by Dr Damian Milton, an autistic British academic, in a 2012 paper published in Disability & Society. Milton is himself autistic, and his perspective as an insider shaped the theory significantly. The underlying idea had been raised by autistic advocates before 2012, but Milton provided the theoretical framework and vocabulary that enabled research to test it directly.
What is the difference between the double empathy problem and theory of mind?
Theory of mind — and its related concept, mind-blindness — proposes that autistic people have a specific cognitive deficit in understanding other people’s mental states. The problem is located in the autistic brain. The double empathy problem challenges this by arguing that the difficulty is an interaction problem: when two people have different neurotypes, both struggle to model each other’s perspective accurately. It does not deny that real social difficulties exist; it reframes where they come from and who is responsible for them.
What is autistic hyper-empathy?
Autistic hyper-empathy refers to the experience of feeling others’ emotions intensely — often involuntarily, and sometimes to an overwhelming degree. Rather than being emotionally detached, many autistic people find themselves absorbing the emotional states of people around them with very little filter. This can manifest as extreme distress at others’ pain, difficulty being in emotionally charged environments, or shutdown in response to emotional overload. Research by Hume and Burgess (2021) found this experience is commonly self-reported by autistic people and directly contradicts the empathy deficit narrative. What looks like emotional withdrawal is often protection from too much, not too little.
What does the research on autistic communication actually show?
Studies show autistic adults communicate just as effectively with other autistic adults as non-autistic adults do with each other. The breakdown occurs specifically in mixed-neurotype interactions. A qualitative companion study found that autistic adults consistently describe within-neurotype interaction as more comfortable, less exhausting, and more mutually comprehensible than cross-neurotype interaction. Together, these findings support the double empathy model: the difficulty is bidirectional and arises from neurotype difference, not from a deficit in autistic communication itself.
How does the double empathy problem relate to autistic masking?
Masking — suppressing natural autistic communication styles to appear more neurotypical — is largely a response to the cross-neurotype mismatch the double empathy problem describes. When autistic communication is consistently read as awkward or confusing by non-autistic people, the pressure to conform is enormous. The double empathy problem reframes that pressure: it is not evidence that autistic communication is wrong. It is evidence that it is different. The masking loop is self-reinforcing: the environment creates the pressure that produces the mask, and then uses the mask’s existence to justify the original demand.
Does the double empathy problem mean non-autistic people are at fault?
Not in a simple moral sense. The double empathy problem is a structural observation about cross-neurotype communication, not an accusation. Non-autistic people forming negative first impressions of autistic people are not necessarily acting with malice — the judgments appear to be rapid and largely unconscious. What the research suggests is that the responsibility for cross-neurotype communication difficulties has been assigned almost entirely to the autistic person, when the difficulty is genuinely bidirectional. A more accurate response would involve non-autistic people and institutions sharing that responsibility, rather than placing all adaptation on the autistic side.
Is the double empathy problem accepted by mainstream autism research?
It is increasingly supported by empirical evidence and acknowledged in research literature, though it has not yet fully displaced deficit-based language in diagnostic criteria or clinical training. The 2022 ten-year review by Milton and colleagues concluded that it has moved from theoretical argument to empirically grounded framework. The shift in how autism research frames its questions — from “what is wrong with autistic social cognition” to “how do cross-neurotype interactions work” — reflects its growing influence, even where official criteria have not yet caught up.