Everyone Masks: What Makes the Autistic Version Different

Say the word “masking” in a room of non-autistic people and someone will eventually say: but everyone does that.

They are right. That is the difficult part. The claim is not wrong, it is not usually said in bad faith, and dismissing it does the autistic case no favours. Impression management is one of the most thoroughly documented behaviours in social psychology, it is universal, and any argument that starts by denying it is going to lose.

So this essay concedes the point in full, and then asks the question that actually matters: if everyone masks, what makes the autistic version different?

There are three main differences often attributed. It costs more to run. It buys less belonging. And it fails more often than it works. There is also a fourth thing, which is not a difference in the masking but in what a person concludes from it, and it is the one that does the lasting damage.

Part 1:
Yes, everyone masks

Erving Goffman set this out in 1959. He described social life as theatre: a front stage where we manage the impression others form of us, and a back stage where we do not.¹ It is one of the most cited ideas in sociology, and it was never about a minority. It was about everyone.

Social psychology later put numbers to it. Leary and Kowalski’s review split impression management into two components - the motivation to control how you are seen, and the construction of the impression itself.² Both are ordinary. Both operate in almost every social encounter. Neither is pathological.

The everyday examples are not hard to find.

A man puts on a dress shirt to go to an office where nobody’s work requires a collar. A woman puts on makeup. Neither is deceiving anyone; both are answering a rule. And the rule works. When Etcoff and colleagues showed faces to raters for 250 milliseconds, cosmetics raised judgements not only of attractiveness but of competence, likeability and trustworthiness.³ A quarter of a second is not enough time to think. The performance is doing real work at a level below deliberation.

Dellinger and Williams interviewed women about workplace appearance rules and found makeup tied to assumptions about health, credibility and professional competence - not preference, but a condition of being taken seriously.⁴ Some of their participants resented it and complied anyway. That is worth holding on to, because it is the same structure autistic people describe: a performance you did not choose, in service of an audience you cannot ignore.

I have now used makeup twice, and I should say why rather than leave it sitting there. It is the example everybody reaches for because it is a literal mask - the only one you can watch being applied, and the only one that leaves evidence in the bathroom sink. That makes it easy to cite and easy to misread as a point about women. It is not one.

The male version is the same behaviour with nothing to photograph. Wong and colleagues pooled 78 samples and 19,453 participants and found conformity to masculine norms unfavourably associated with mental health, and more strongly with willingness to seek psychological help - with self-reliance and emotional control among the norms most consistently implicated.⁵ A man holding a flat expression through a bad week, answering “fine” because within the rule no other answer is available, is performing to a code with the same force as any dress standard. He simply does not have to buy anything to do it.

Which is the part worth extracting, because it runs through the whole essay. The masks we can see get noticed, discussed and argued over. The masks that leave no trace go unremarked, which makes them harder to name and correspondingly harder to question. The autistic mask is the second kind.

Underneath all of it sits normative social influence. Deutsch and Gerard separated two reasons people conform: because they believe the group is right, and because they want to be accepted by it.⁶ The second reason has nothing to do with truth. It is about not being excluded. Most of what we call “normal behaviour” is normative in exactly that sense - a set of conventions we perform to remain in the group.

Which produces a small irony worth sitting with. If normal has to be put on, it is not normal. It is a costume with very good marketing.

The phrase people reach for here is fake it till you make it. I have not been able to trace it to a citable source, and I am not going to pretend otherwise.⁷ But the idea it points at is real enough: act the part long enough and the part stops being an act. That is the mechanism. It is also, precisely, where the autistic case diverges.

Recent research has stopped treating masking as autism-specific. Pryke-Hobbes and colleagues compared 285 autistic, 88 non-autistic neurodivergent and 99 neurotypical adults on workplace masking, and found masking widespread in all three groups, for overlapping reasons: to fit in, to avoid judgement, to keep the job, to get promoted.⁸ Miller, Rees and Pearson found the same overlap comparing autistic and non-autistic adults directly.⁹ Van der Putten and colleagues found that adults with ADHD camouflage more than a non-autistic, non-ADHD comparison group - though still less than autistic adults.¹⁰

Ai, Cunningham and Lai put the conceptual case plainly: autistic camouflaging sits inside the general human phenomenon of impression management, and the useful research question is which parts are shared and which are not.¹¹

So the honest starting position is this. Masking is not an autistic invention, autistic people are not the only ones doing it, and the difference is not a wall. It is a gradient. But the gradient is steep, and it is steep in specific places.

Part 2:
The first difference: what it costs to run

The clearest way to understand this comes from a literature that has nothing to do with autism.

Organisational psychologists studying emotional labour distinguish surface acting from deep acting. Surface acting is displaying an emotion you do not feel - the smile held in place over nothing. Deep acting is adjusting the internal state so the display is genuine. Hülsheger and Schewe’s meta-analysis of 95 studies found that surface acting has substantial relationships with impaired wellbeing, with correlations between .39 and .48, and negative relationships with job attitudes. Deep acting showed only weak relationships with impaired wellbeing, and positive relationships with performance.¹²

Same outward behaviour. Opposite internal economics. What separates them is whether the performance has been absorbed or is being held up by hand.

For most non-autistic people, most social presentation is absorbed. Bargh and Chartrand’s review of automaticity argued that a great deal of social behaviour runs without conscious direction at all.¹³ Their companion work described what they called the chameleon effect: people unconsciously mimic the posture, mannerisms and expressions of whoever they are talking to, and that mimicry measurably increases how much the other person likes them.¹⁴

Read that twice. The literature’s own word for unconscious neurotypical social mimicry is chameleon, and it is described as social glue. The same word gets used as an accusation against autistic people who do the identical thing on purpose.

This is where the “I don’t leave the house without putting my face on” example earns its place. What that phrase describes is a performance that has been absorbed into identity. The face is not experienced as a costume; it is experienced as the person. It has become deep acting. There is no running cost because there is nothing being held up.

The male equivalent has no object attached to it, which is exactly why it goes unremarked: the automatic “fine”, the flattened register, the enthusiasm held a notch below where it actually sits. Absorbed the same way, running at the same low cost, and invisible because there is nothing to see.

The autistic version does not absorb.

Livingston and Happé’s framework for compensation in autism describes it directly: what neurotypical people achieve implicitly, some autistic people achieve explicitly, through reasoning - and higher intelligence assists that process.¹⁵ Explicit reasoning is the expensive kind. It runs in working memory, in real time, alongside the actual conversation. Hull and colleagues’ interviews with autistic adults describe exactly that: rehearsed scripts, borrowed mannerisms, deliberate eye contact, monitored posture - and exhaustion as the near-universal consequence.¹⁶

Scheeren and colleagues put this on a clock. Using ecological momentary assessment, they sampled 87 autistic adults repeatedly through ordinary days and found that masking rose when other people were present, rose further when the others were non-autistic, and that higher masking at a given moment went with higher perceived stress at that same moment.¹⁷ Not remembered stress. Concurrent stress, measured while it was happening.

The cost also shows up in the body. A co-twin control study measured camouflaging against hair cortisol concentration - a biological record of long-term stress - in a neurodiverse twin sample of 315 people, 69 of them autistic. Camouflaging was associated with raised cortisol, particularly in the autistic and adult subsamples, though in the full sample it was not associated with self-rated stress symptoms.¹⁸ That last clause matters: the body was registering something the questionnaire was not.

Now the part where I have to complicate my own argument, because the evidence does. Three complications, and none of them are small.

Autistic masking is often not deliberate. In the Pryke-Hobbes study, autistic and neurodivergent participants repeatedly described masking as habitual and largely automatic - “second nature”, something done for decades without noticing. One participant put it as bluntly as it can be put: by the time you reach forty-seven and only learned you were autistic at forty-five, masking is such a habit as to really not be a choice.⁸

Neurotypical people are not exempt from the cost. In the same study, neurotypical participants described masking as too much effort to sustain, described feeling like outsiders because of it, and described a vicious circle of self-monitoring.⁸

The direction of causation is genuinely unsettled. The strongest design here is the co-twin comparison, and it supports the causal reading: among both identical and non-identical twin pairs - which control for shared genetics and upbringing - the twin who camouflaged more reported lower quality of life than their co-twin.¹⁹ But a longitudinal study of 332 autistic adults measured two years apart found the opposite of what everyone expected. Camouflaging and mental health difficulties were associated at baseline, in line with all the cross-sectional work. Over the two years, though, higher initial camouflaging predicted a small decrease in mental health difficulties, and lower initial camouflaging a slight increase.²⁰ The effects were small and one study does not overturn a literature. But it is a real result, from a good design, pointing the other way, and anyone making confident causal claims about masking - including me - has to hold it.

So the clean version of the claim - they do it unconsciously and cheaply, we do it consciously and it destroys us - does not survive contact with the data. The version that does survive is narrower and better. Habit can start the mask without any decision being made; it still costs to run once started, because the execution is explicit compensation rather than absorbed behaviour. Cook and colleagues’ systematic review found exhaustion reported as a consequence of camouflaging with striking consistency across the literature.²¹ A habit you cannot stop and cannot afford is worse than a choice, not better.

Part 3:
The second difference: it does not buy belonging

The entire point of masking is to belong. That is the transaction: I will perform your conventions, and in exchange I get to be inside the group.

It is worth being precise about what is being bought. Baumeister and Leary’s review established belonging as a fundamental human need rather than a preference - a drive for frequent, non-aversive interaction inside a stable, continuing relationship, with its frustration producing measurable harm.²² Masking is not vanity and it is not social climbing. It is a bid for something the nervous system treats as a requirement.

For autistic people the transaction routinely does not clear.

Milton and Sims analysed autistic adults’ own accounts of wellbeing and belonging and found “living with the consequences of an othered identity” as one of the four organising themes.²³ Not a failure to want connection - a persistent experience of standing outside it. Umagami and colleagues’ systematic review of loneliness in autistic adults found loneliness consistently reported at elevated levels, alongside a desire for connection that the loneliness did not reflect a lack of.²⁴

That last point is the one most often missed. The outsider feeling is not solitude, and it is not introversion. It is the experience of being in the room and not in the group.

At work

The workplace is where this is most visible, because the workplace supplies the contact and withholds the belonging.

The Pryke-Hobbes participants explain the mechanism in their own words. Friendships formed behind the mask were described as friendships with the persona rather than the person. Participants reported that no one really knew them. One described working alongside the same people for seventeen years, watching others form lifelong friendships, and never becoming close to any of them. Another described feeling like a Russian doll and wondering whether anything of the real self was left when the layers came off.⁸

None of those people were isolated. They had colleagues, meetings, invitations, teams. What they lacked was not contact. It was recognition.

And the workplace has a specific cruelty about it: masking there is the least optional and the most rewarded. Participants described masking as the condition of getting hired, being promoted, and being trusted with responsibility - one said that projects and promotions rest on popularity and seeming normal even in companies that claim otherwise.⁸ So the environment pays for the performance and withholds the thing the performance was for. That is not a contradiction the person can resolve from the inside.

Socially

Outside work the shape is different and the outcome is similar.

Social masking is less scripted and, several autistic participants reported, harder for that reason. Professional roles come with explicit rules; friendship does not. One said that with the right knowledge you can basically learn the script for a work role. Another said that if friends provided a written contract and a job description, they might be better at friendship.⁸

So the social version is a performance with no rulebook, run for an audience whose expectations are unstated, in the one domain that is supposed to be voluntary and restorative. Invitations get declined. The decline reads as disinterest. Over years, the circle contracts, and the contraction is read - by everyone, including the person - as evidence about their character rather than about the cost of the room.

Why the transaction fails

The failure is structural, and it has a specific mechanism: what gets accepted is the performance, and acceptance does not transfer.

It is not that the acceptance is fake. The colleagues mean it. It is that they are accepting a different person, and you know it, continuously, while it is happening. Being liked for something you are not is a lonelier experience than not being liked at all, because there is nobody to tell.

Cage, Di Monaco and Newell tested the part that matters here. Surveying 111 autistic adults, they found that acceptance - both from others and personal acceptance of being autistic - significantly predicted depression, and that acceptance from others also predicted stress. Camouflaging, in the same study, related to higher rates of depression.²⁵ So the operative variable is not how much social contact a person has. It is whether they are accepted as what they are. Masking, by design, makes that impossible to find out.

Botha, Dibb and Frost’s interviews explain why the stakes are so high. Their participants described autism as central to identity, integral to who they are - and described society conferring negative meaning onto autism, and therefore onto them.²⁶ Concealing autism is not concealing a habit. It is concealing the thing you understand yourself to be.

Pearson and Rose reframe masking as a response to stigma rather than a personal strategy, and argue that the “choice” to mask is largely an illusion when the alternative is being treated as lesser.²⁷ Perry and colleagues came at it through social identity theory and found the same shape: camouflaging as a response to a devalued group identity.²⁸

And the price is measurable at the top end. In the interpersonal theory of suicide, thwarted belongingness is a specific construct - the unmet need to belong - which, combined with perceived burdensomeness, is proposed to produce the most dangerous form of suicidal desire.²⁹ Cassidy and colleagues took that framework and applied it to camouflaging directly, finding camouflaging associated with thwarted belongingness and with lifetime suicidality.³⁰ The thing masking is supposed to buy is the thing its heaviest users most conspicuously lack, and the shortfall has a body count.

Botha and Frost make the wider point that this harm follows the shape of minority stress: it comes from social position, stigma and concealment rather than from the neurology itself.³¹ That is not a small distinction. It means the outsider feeling is not a symptom. It is a response.

I want to keep one caution attached here. A neurotypical participant in the Pryke-Hobbes study also said masking made them feel like an outsider, and the seventeen-years-without-a-friendship quote came from a neurotypical participant too.⁸ The experience is not exclusive. What differs is dose, duration, and whether there is anywhere the mask comes off.

And there is a further complication that cuts against the simple deficit reading. Milton’s double empathy argument holds that the breakdown between autistic and non-autistic people is mutual - two groups failing to read each other, not one group failing to read the other.³² Crompton and colleagues gave that empirical support by showing that information transfer between autistic people is as effective as between non-autistic people, and degrades only in mixed pairs.³³ The isolation is not a straightforward property of the autistic person. It is a property of the pairing.

Part 4:
The third difference: it often does not work

Here is the part that gets least attention and deserves the most.

Autism is defined diagnostically by persistent deficits in social communication and social interaction - including deficits in reading and using nonverbal communicative behaviours, and in adjusting behaviour to suit varying social contexts.³⁴ Set that next to what masking requires. To perform a social convention correctly you must first read it correctly: the register, the timing, the volume, the amount of eye contact this particular person expects, the point at which enthusiasm becomes too much.

The mask is built out of exactly the information the disability restricts access to. So the performance is assembled from guesswork, and a proportion of it comes out wrong.

The wrongness is not subtle and it is not slow. Sasson and colleagues showed observers brief clips of autistic and non-autistic adults and found autistic people rated less favourably across a range of traits, with reduced intentions to interact with them - and the judgements formed within seconds.³⁵ The finding that matters most is what happened when the audiovisual cues were removed: with only the content of what was said, the bias disappeared. Observers were not reacting to what autistic people said. They were reacting to how it was delivered.

Then Belcher and colleagues ran the study that tests the mask directly. Eighty autistic and non-autistic adults were recorded in conversation with someone unaware of their diagnostic status. Ten-second clips went to 127 non-autistic raters. Autistic participants were rated more poorly - and camouflaging intent did not predict first impressions at all.³⁶

That is the finding to sit with. The effort does not buy the outcome. People who were actively trying to mask were not perceived any better than people who were not.

Which produces the central irony of the whole business. Masking exists to make you unremarkable. Performed with imperfect access to the rules, it makes you more conspicuous, not less - because a convention performed slightly wrong reads worse than the convention simply absent. There is a specific quality of wrongness to an off-timed laugh or a held-too-long stretch of eye contact that plain difference does not have. The mask can push you further outside the group it was built to get you into.

And the effort does not stop when the encounter does. The Pryke-Hobbes participants described sustained rumination afterwards - replaying the day, hunting for the moment it went wrong, unable to relax at home because the review was still running.⁸ The cost is not only the performance. It is the audit.

Meanwhile, the thing that does measurably improve how autistic adults are received is the opposite of the mask. Sasson and Morrison found that first impressions of autistic adults improved when the diagnosis was disclosed, and improved further when the observer knew something about autism.³⁷ What works is being known. What we spend our lives doing is the reverse.

Part 5:
A different mask for every room

Masking is not one performance. It is several, and they are not interchangeable.

Cage and Troxell-Whitman asked 262 autistic adults about camouflaging across contexts - work against family, formal against informal - and found something important about the pattern rather than the amount. Both camouflaging heavily across all contexts and switching between camouflaging in some contexts and not others were associated with poorer mental health.³⁸ The switching is its own cost, separate from the volume.

Scheeren’s momentary data shows the switching happening in real time: masking lower when alone, higher with others present, higher again when the others are non-autistic.¹⁷ The mask is not a fixed object. It is being re-selected continuously against the room.

The work mask is the heaviest and the least optional. Pryke-Hobbes participants described a specific absence of choice at work - no control over which colleagues you get, no ability to leave when it becomes too much, and income riding on the performance. One participant said the risks of being different at work change your life because they are linked to your income.⁸

The home mask is stranger and less discussed. For some people home is genuinely the place the mask comes off - the small set of people who already know, where being uncommunicative is permitted. For others it does not come off at all. A mask worn with a partner, with children, with parents, is worn without any of the compensations work provides: no defined role, no script, no end of shift, and an audience that has known you for decades and would notice a change. And it carries a cost the work mask does not, because the relationship it protects is the one that was supposed to be the safe one. The Pryke-Hobbes participants described exhaustion from workplace masking spilling into home relationships until there was nothing left for the people there - which is its own kind of home mask, the one where you perform having capacity you do not have.

Then there is the collision problem, and everyone who lives this knows it before it is explained.

Two masks cannot run at once. So the wedding where colleagues and family are both invited becomes something to avoid. The work social event your partner is welcome at becomes something to decline. Not because either group is unwelcome, but because the two performances are mutually exclusive and there is no third one that satisfies both. The result reads to everyone else as aloofness, or as hiding something. It is neither. It is a scheduling problem in a system with no spare capacity.

And the people who see more than one version draw the obvious conclusion. The Pryke-Hobbes participants reported being seen as two-faced, as a charlatan, as dishonest - and reported feeling those things themselves.⁸ The word most commonly attached from outside is chameleon, which brings us back to Chartrand and Bargh, who used the same word to describe automatic neurotypical mimicry and called it social glue.¹⁴ Do it without noticing and it is rapport. Do it deliberately, and imperfectly, and it is duplicity.

Lai and colleagues’ work on quantifying camouflaging is a useful reminder here that this is measurable, individually variable, and present in both autistic men and women.³⁹ It is not a character trait. It is a workload.

Part 6:
What the rooms cost: the diagnostic record

A person who presents differently in every context, whose presentation is effortful and unstable, and who collapses when the audience leaves, looks like something specific from the outside. It usually gets a name long before it gets the right one.

Fusar-Poli and colleagues reviewed the clinical records of 161 adults who received a first autism diagnosis at two Italian university centres. The median age at diagnosis was 23. The median gap between first contact with a mental health professional and the autism diagnosis was 11 years. Only 33.5% had never received a psychiatric diagnosis before - meaning roughly two thirds arrived carrying at least one other label, and a substantial minority carried several. One woman had six. One man had eight.⁴⁰

The breakdown is worth stating in full, because vague claims about misdiagnosis do more harm than precise ones. Prior diagnoses in that sample were intellectual disability (most common, and disconfirmed by testing in nine of 35 cases), schizophrenia and other psychotic disorders at 16.1%, personality disorders at 14.9% - mostly personality disorder not otherwise specified, with smaller numbers of schizoid, borderline and schizotypal - depression at 13.7%, anxiety disorders, obsessive-compulsive disorder and conduct disorders at 7.5% each, ADHD at 6.8%, gender identity disorder at 3.1%, and bipolar disorder and language disorder at 2.5% each.⁴⁰

And then the number that matters most. When the assessing team checked whether those prior diagnoses held up as genuine co-occurring conditions, they held up for 48.6% of the people who had received them. More than half of the labels those adults had been carrying, some for over a decade, were wrong.⁴⁰

Kentrou and colleagues approached it from the other side, asking 1,211 autistic adults in the Netherlands which of their prior psychiatric diagnoses they considered a misdiagnosis. Just under a quarter - 24.6% - reported at least one. Personality disorders were the most frequently named, ahead of anxiety disorders, mood disorders, chronic fatigue and burnout-related disorders, and ADHD. Women reported perceived misdiagnosis nearly twice as often as men: 31.7% against 16.7%.⁴¹

Two independent methods, two countries, one label at the top of the list.

I want to be careful about what this does and does not show. It is not evidence that personality disorders are fictional, and it is not evidence that clinicians are careless. Dudas and colleagues compared 624 people with autism spectrum conditions, 23 with borderline personality disorder and 16 with both against 2,081 controls, and found that the borderline group had genuinely elevated autistic traits and an elevated drive to systemise.⁴² The overlap is real at the level of the traits themselves, not just at the level of the paperwork. A clinician looking at unstable self-presentation, identity disturbance and interpersonal difficulty is not hallucinating.

Nor is masking the only reason autism gets missed. Diagnostic criteria built on children, clinicians trained on a young-boy prototype, and co-occurring conditions that absorb attention all contribute, and I have written about those elsewhere.

But masking belongs on the list, and the shape of the misdiagnosis pattern is suggestive. Identity disturbance is a personality disorder criterion. A person running a different persona in every context, unable to say which one is real, describing themselves as a Russian doll with nothing at the centre, is presenting identity disturbance - correctly. What is wrong is not the observation. What is wrong is the inference about its cause.

One gap I should name rather than paper over. Popular accounts often add dissociative identity disorder - “multiple personality” - to this list, and I could not find peer-reviewed prevalence data supporting it in autistic adults.⁴³ I have left it out. Bipolar disorder, likewise, sits at 2.5% in the Fusar-Poli sample - present, but not a headline number, and I am not going to inflate it into one.

The consequence for the person is not only clinical. A decade of being told what is wrong with you, by people with training and authority, where each answer is confidently delivered and none of them work, does something to how much you trust your own account of your own life. By the time the right answer arrives, the habit of not being believed is well established - including by yourself.

Part 7:
The other thing that breaks: work

The second measurable outcome is employment, and it is the one where the arithmetic of surface acting becomes visible.

Bury and colleagues followed 2,449 autistic adults across eight annual waves of the Netherlands Autism Register and identified four employment trajectories. The largest, at 1,189 people, was stable unemployment. Stable employment accounted for 801. A further 134 started with a high probability of employment that declined across the eight years.⁴⁴ Just under half the sample was not working, and did not start.

The Pryke-Hobbes sample shows the same gap inside a single study: 37.6% of autistic participants in full-time employment against 80.3% of neurotypical participants, and 44.5% satisfied with their employment status against 85.2%.⁸

Baldwin, Costley and Warren surveyed 130 Australian adults and found what the pattern looks like up close - people with the capacity and the willingness to work, facing significant labour market disadvantage and working below their qualifications, with little understanding or support in the workplace.⁴⁵

Masking is not the sole explanation for any of this. Discrimination, inaccessible recruitment, sensory environments and inflexible hours all operate independently. But the participants describe the mechanism themselves, and it is a mechanism of depletion rather than incapacity.

One said that masking sacrificed their abilities - that they heard less, missed things, burned more energy, and could not use their mind in the ways they knew they could. Another said that if they could not fit in, they found the job harder and harder to do. Participants reported extended sick leave, and reported leaving jobs permanently.⁸ What is being described is not someone who cannot do the work. It is someone spending the capacity the work needs on the performance surrounding the work.

That has a name. Higgins and colleagues, working with autistic adults as experts by lived experience, defined autistic burnout as a severely debilitating condition whose onset is preceded by fatigue from camouflaging or masking autistic traits, from interpersonal interaction, from cognitive overload and from unaccommodating sensory environments.⁴⁶ Masking is the first item on that list. Raymaker and colleagues’ earlier definition describes the same collapse - internal resources exhausted beyond measure, with no clean-up crew.⁴⁷

Surface acting predicts exhaustion, and exhaustion predicts leaving. The organisational literature established that link in populations who mask for eight hours and then stop.¹² Autistic employment outcomes are what that relationship looks like when the shift never ends.

I could not find a peer-reviewed figure for average job tenure in autistic adults, which is the statistic this section most obviously wants.⁴⁸ The figures circulating online do not trace to a study I can check, so I have not used them.

Part 8:
The verdict you reach without a diagnosis

Everything above describes a mechanism. This part is about what a person concludes from living inside it for forty years without being told it exists.

The only explanation available

Every failed interaction demands an explanation. The mind does not tolerate a pattern of unexplained failure; it will supply a cause whether or not it has the information to do so correctly.

Abramson, Seligman and Teasdale set out the framework in 1978. When people explain negative events, the explanations vary along three dimensions: internal or external, stable or unstable, global or specific. The combination that produces the most damage - depression, and the generalisation of helplessness across situations - is internal, stable and global. Not that went badly, but I am the kind of thing that goes badly, and always will, everywhere.⁴⁹

Now consider what is actually available to an undiagnosed autistic adult trying to explain a life of these events. The conversations that curdled for no visible reason. The jobs that ended without a stated cause. The friendships that thinned. The relationships that failed. From the inside there is no visible mechanism, because the mechanism is a mask nobody has named and a social-information deficit nobody has diagnosed. The one variable present at every single failure is the person themselves.

So the attribution is internal, stable and global, and it is arrived at rationally. Given the evidence available, it is the correct inference. It is simply drawn from an incomplete dataset, and nobody has supplied the missing column.

Which makes it shame, not guilt

The distinction matters more than it sounds.

Tangney, Stuewig and Mashek summarise decades of work separating the two. Guilt is about behaviour - I did a bad thing - and it motivates repair, apology and change. Shame is about the self - I am a bad thing - and it motivates concealment, withdrawal and avoidance of the person you failed in front of.⁵⁰

For an undiagnosed autistic masking failure produces shame rather than guilt, because there is no identifiable behaviour to attach the guilt to. You cannot apologise for a thing you cannot name. What remains is the self.

And now the loop closes, because the behavioural response to shame is to hide. The response to being wrong, when you do not know what is wrong, is to conceal more of yourself and perform harder. Which is more masking, which produces more failure, which produces more shame.

That is not a metaphor for a feedback loop. It is one, and each pass tightens it.

What the evidence says about shame specifically

This is not only a vague hypothesis. Riebel and colleagues surveyed 689 autistic adults on self-stigma, self-compassion, camouflaging and depression, and found that self-stigma raised depressive symptoms through shame - shame as the mediating pathway, not an incidental feeling alongside it.⁵¹ Botha and Frost’s minority stress account puts the same finding in its social context: the damage tracks stigma, concealment and social position rather than the neurology.³¹

The repair is slow and it starts late. Corden, Brewer and Cage surveyed 151 autistic adults on identity, self-esteem and wellbeing after diagnosis, and found that more time elapsed since diagnosis related to less dissatisfaction with autistic identity.⁵² Read that the other way round and it is bleak: the work of not despising yourself for being autistic cannot begin until someone tells you that you are, and then it takes years.

Which is the case for diagnosis stated without sentiment. A diagnosed child gets handed an external, specific, unstable attribution: this is a known thing, it has a name, here is what helps. An undiagnosed adult had four decades of the internal, stable, global one. The diagnosis does not remove the accumulated damage. It stops the accrual.

Trauma, and a distinction that gets collapsed

The obvious next question is whether decades of this amounts to trauma in the clinical sense rather than the colloquial one. Answering it needs a distinction that is routinely flattened, including in a good deal of autism writing.

Post-traumatic stress disorder (PTSD) follows a discrete threatening or horrific event, or a small number of them. Complex post-traumatic stress disorder (CPTSD) - recognised in the ICD-11, and not recognised in the DSM-5 at all - follows prolonged or repeated events from which escape is difficult or impossible. But the difference is not only in the cause. CPTSD requires everything PTSD requires plus a second symptom cluster, disturbances in self-organisation, made up of three parts: affect dysregulation, disturbances in relationships, and negative self-concept.⁵³

That third part is why this section exists. Negative self-concept - persistent beliefs about oneself as worthless, defeated or diminished - is a diagnostic criterion of CPTSD and is not a criterion of PTSD. It is, almost word for word, the verdict the attribution problem above produces.

So the construct that fits the pattern this essay describes is the complex one, on both counts. The events are not discrete or extreme; they are ordinary, repeated, low-grade and inescapable. And the damage lands in self-concept rather than only in threat response. Which also means that PTSD evidence, however plentiful, is not evidence for this claim. It sits next to it.

Here is what the literature does support.

Rumball, Happé and Grey found autistic adults reporting a wide range of life events as traumatic - importantly, many that would not meet the standard diagnostic threshold for a traumatic event at all - and found trauma symptomatology following those non-qualifying events.⁵⁴ The valuable part of that finding is not the headline rate. It is the demonstration that chronic, sub-threshold, socially grounded adversity - exclusion, bullying, sustained invalidation - registers as traumatic in this population. That is the exposure type the complex construct was built around.

Quinton and colleagues followed a nationally representative cohort of 2,232 twins in England and Wales and found children with higher autistic traits significantly more likely to meet criteria for PTSD by age 18 - though their higher rate of trauma exposure was statistically accounted for by co-occurring socio-economic disadvantage, so exposure and outcome are not one story.⁵⁵ That is a PTSD finding, and I am labelling it as one rather than letting it stand in for something it does not measure.

Here is what the literature does not support.

A 2025 systematic review and meta-analysis pooled diagnosed PTSD prevalence in autistic people at 2.06% in adults - broadly the general population rate, and wildly at odds with the 32-45% that screening instruments produce.⁵⁶ Both figures cannot be describing the same thing. Either PTSD is being systematically missed in autistic adults, or screening instruments are counting autistic traits as trauma symptoms. Note that this is a measurement failure in the simpler of the two constructs.

For complex post-traumatic stress disorder in an autistic sample, I could not find a prevalence estimate at all.⁵⁷

And there is good reason to expect that gap to be hard to close, because it is the personality disorder problem one storey up. Take the three disturbances in self-organisation in turn. Affect dysregulation is describable as autistic emotional response, as post-masking depletion, or as complex trauma. Disturbances in relationships is describable as autistic social difference, as the wreckage of masked friendships, or as complex trauma. Negative self-concept is describable as internalised stigma, as the shame this essay has just traced, or as complex trauma. Every criterion has at least three plausible owners. Sarr and colleagues convened 106 international clinical experts in a three-round Delphi study specifically to establish how to distinguish autism from attachment disorders, CPTSD and emotionally unstable personality disorder.⁵⁸ One do not run a study like that unless the distinction is failing in practice.

So the position I will defend is this and no more. The mechanism is well evidenced: chronic invalidation, an internal-stable-global attribution drawn from incomplete information, and shame as a measured mediator to depression. The exposure pattern matches the complex construct rather than the simple one, and autistic adults are documented as experiencing exactly that kind of sub-threshold, repeated adversity as traumatic. What is not established is any prevalence figure for CPTSD in this population, and I am not going to assert one because it matches my own experience. What I will say is that the profession has noticed the overlap and is actively working on it, and that anyone in this position is entitled to be assessed for both rather than sorted into whichever construct the clinician happens to specialise in.

Why this is the part that lasts

The other three differences are costs you pay while the mask is on. This one is what you are left holding afterwards.

Exhaustion resolves with rest. Employment can be rebuilt. A diagnosis can be obtained at fifty, and the misdiagnoses corrected. But a settled, decades-old conviction that you are fundamentally wrong does not lift when someone hands you the correct explanation, because it was never held as a belief you could argue with. It was held as a fact about the world, laid down before you had the capacity to question it, and confirmed by every failure since.

That is why the “everyone masks” conversation is worth having properly rather than winning. The people saying it are not describing this. They have never had to explain a lifetime of unexplained failure with no data except themselves.

Part 9:
So what is the difference

Everyone masks. That is true, and conceding it costs nothing, because the concession is where the argument starts.

Three things separate the autistic version.

It never converts. Neurotypical presentation mostly becomes the person - deep acting, absorbed, cheap to run. The autistic version stays surface acting, held up by explicit reasoning, in real time, with the cost landing in the same moment and registering in the body.^12151718 Habit may remove the decision, but it does not remove the bill.

It does not buy belonging. The acceptance goes to the performance, and the person knows it while it is happening. Belonging is a need rather than a preference, and what the heaviest maskers report is not membership but thwarted belongingness - the specific loneliness of being surrounded and unrecognised.^22232530

It fails more often than it works. The mask is assembled from social information the disability restricts access to, so it comes out wrong in ways that make the wearer more conspicuous, not less. Camouflaging intent does not predict how people are actually received.³⁶ What does improve reception is disclosure - being known rather than concealed.³⁷

And then there is what those three produce together, in a person who has not been told any of it. Repeated unexplained failure, attributed inward because there is nowhere else to attribute it, hardening into shame, which prescribes more concealment, which produces more failure.⁴⁹⁵⁰⁵¹ That is the part that outlasts the diagnosis.

None of that is an argument for unmasking on command. Pearson and Rose are right that the choice is largely an illusion when the alternative is being treated as lesser, and the participants who described being bullied almost to extinction when they masked less were not describing a hypothetical.²⁷⁸ Telling someone to stop masking without changing what happens when they do is asking them to absorb a cost you are not paying.

The instruction that follows from the evidence is not aimed at autistic people at all. Autistic adults are already doing the expensive, unreliable, self-erasing version of a task everyone else does automatically and for free. The only intervention with measured evidence behind it moves the effort somewhere else: people who know something about autism form better impressions of autistic people, without the autistic person having to perform anything.³⁷

But everyone masks is a true sentence and half an argument. The half it leaves out is that for most people the mask becomes the persona, buys the acceptance it was meant to buy, works most of the time, and - importantly - has a stop: at the door, with the few who already know, or by becoming cheap enough that it stops registering as masking at all. The autistic version is perpetual; it is concentrated effort, spending the capacity that the work and the feelings needed on the performance wrapped around them; it is often so wrong that it creates the very ostracism it was built to prevent; and it has no reliable stopping point. Some of us find it at home, where autism is known and accepted. Some of us have looked everywhere and never found it.

⁂

---