The Loneliness Snowball: How Autistic Isolation Feeds Itself

There is a sentence autistic people hear a lot: but you like being on your own.

It is not a stupid thing to say. It is often true. I do like being on my own; I need it, and I get less done and feel worse without it. The problem is what the sentence is being used for. It is used to close the question - to explain the empty calendar as a preference, and to move on.

This essay is about what that explanation misses. Autistic loneliness is not a taste for solitude. It is a loop, and the loop has a direction. Each turn of it makes the next turn more likely, which is why it does not stay the size it started at.

I have called it a snowball because that is what it does. It is small at the top of the hill and it is not small at the bottom.

Part 1:
What is actually being measured

First, the thing itself, because loneliness gets used for several different states and the confusion does real damage.

Being alone is a fact about the room. Loneliness is a fact about the gap between the connection you have and the connection you want. You can be alone and not lonely. You can be surrounded and lonely. They are not the same variable and they do not have to move together.

But there is a third state, and it is the one autistic people describe most often. Not alone. Not simply lonely. Present, but an outsider.

I will describe it the way it actually runs, because I have never found a tidier way to say it. I am in a supermarket, or an office, or somebody’s kitchen at a party, and I am observing. Not participating - observing, the way you would watch a species. I can follow what everybody is doing and why. I can explain the rules back to you afterwards, sometimes better than the people obeying them could. What I cannot do is be inside it. And underneath that sits something harder to admit, which is a low, persistent doubt about whether I am quite the same kind of thing as the people I am watching.

I want to be careful about how that reads, because there is an obvious misreading and it is not a kind one. It is not superiority. Standing outside a group you would very much like to be in is the opposite of looking down on it. But I have been accused of coldness often enough to check myself for it before anybody else gets the chance, which is its own small tax.

Milton and Sims analysed autistic adults’ own accounts of wellbeing and belonging and found “living with the consequences of an othered identity” among the four organising themes. ¹ Not a failure to want connection. A standing experience of being outside the thing you are physically inside.

And the not-quite-human part is not purely a feeling, which is the detail that stopped me short when I found it. Cage, Di Monaco and Newell surveyed 361 non-autistic people and found evidence of dehumanisation towards autistic people - specifically, the denial of the traits regarded as uniquely human. ² So an autistic adult who reports not being certain they are fully human is not manufacturing a symptom. They are reporting something that is being done to them, and reporting it accurately.

That distinction runs through the rest of this essay. Most of what looks like an autistic feeling about the world turns out, on inspection, to be an accurate reading of it.

Umagami and colleagues reviewed 34 studies of loneliness in autistic adults and found it consistently elevated, alongside a set of factors that raise it: anxiety, depression, suicidal ideation, negative experiences and learned helplessness, a lack of autism understanding and acceptance, sensory avoidance, camouflaging, and unemployment. ³ Read that list again. Almost none of those are properties of the autistic person in isolation. Most of them describe a relationship between a person and an environment.

Muris and colleagues put a size on the gap. Reviewing the field, they report loneliness in autistic populations at roughly three to four times the rate found in non-autistic comparison groups - and, at the same time, a stronger stated preference for solitude. ⁴ They call the combination the solitude-loneliness paradox, and I will come back to it, because it is the hinge this whole essay turns on.

The reason any of this matters is that belonging is not a preference. Baumeister and Leary’s review established the need to belong as a fundamental human motivation - a drive for frequent, non-aversive contact within a stable, continuing relationship, whose frustration produces measurable harm. ⁵ It is a requirement, not a hobby. Which means that an autistic adult with an empty calendar is not having their needs met by the emptiness; they are running a deficit in something the nervous system treats as necessary.

And the deficit is not a lack of wanting. Quadt and colleagues asked 209 people about sensory differences, loneliness and mental health, then recorded autistic adults talking to each other about loneliness directly. Sensory differences were associated with higher loneliness and worse mental health in both groups, more strongly in the autistic one. The first-hand accounts, the authors note, refute the stereotype that autistic adults are not socially motivated. ⁶ The title of the paper is a participant’s own words: I’m trying to reach out, I’m trying to find my people.

Jaswal and Akhtar made the general version of that argument. A great deal of influential autism theory, they wrote, rests on the assumption that certain autistic behaviours - low eye contact, infrequent pointing, stereotypies, echolalia - indicate a lack of social interest. That assumption is contradicted by autistic people’s own testimony, and there are alternative explanations for every one of those behaviours. ⁷

So the starting position is this. The want is present. The connection is not. The gap between them is the thing being measured, and it is three to four times the size it is for everyone else.

I should say plainly that the loop I am about to describe is not my discovery. Schiltz and Lord published an Autism and Loneliness Theoretical Framework setting out contributing factors to both the emergence and the persistence of loneliness in autistic adults. ⁸ The separation of those two words is the whole point, and it is theirs. What follows is my version of the same argument, written from the inside.

Now, how does the gap get made?

Part 2:
The first push: you are outside before you have done anything

The snowball does not start with a decision. It starts with a judgement made about you, by other people, faster than either of you can think.

Sasson and colleagues showed observers brief clips of autistic and non-autistic adults. The autistic adults were rated less favourably across a range of traits, and observers reported reduced intentions to interact with them. The judgements formed within seconds. The critical part is what happened when the audiovisual cues were stripped out and only the content of speech remained: the bias disappeared. ⁹ Observers were not responding to what autistic people said. They were responding to how it arrived.

That is the first push, and it happens before any conversation has taken place. It is not a consequence of autistic behaviour in any meaningful sense; it is a consequence of autistic presence.

Milton’s double empathy argument reframes what is going on. 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. ¹⁰ Crompton and colleagues supplied the experimental version, showing that information passes between autistic people as effectively as between non-autistic people, and degrades only in mixed pairs. ¹¹ Morrison and colleagues took it into real conversation: autistic adults paired with autistic partners had better interaction outcomes than autistic adults paired with non-autistic ones. ¹² Mitchell, Sheppard and Cassidy trace the developmental and mental health consequences of that mismatch. ¹³

The mismatch matters here for one specific reason. If the failure were located in the autistic person, then it would follow you everywhere. It does not. It appears in mixed pairs and thins out in matched ones. That is a property of the pairing, and the pairing is something the world decides, not you.

Then the push gets harder.

Trundle and colleagues pooled the victimisation literature and found a prevalence of 44% across autistic samples, with bullying specifically at 47%. ¹⁴ Nearly half is not a subgroup. It is the ordinary case.

It also does not stop at the school gate. Weiss and Fardella compared 45 autistic and 42 non-autistic adults and found the autistic group reporting more teasing and emotional bullying in adulthood, and more sexual contact victimisation - with no difference between the groups on perpetration. ¹⁵ It is a small sample and I am not going to build much on it, but the direction is consistent with everything else here: more done to us, not more done by us.

And then it reaches money, which is where exclusion stops being social and becomes structural.

In Australia, the Bureau of Statistics’ 2022 disability survey found the unemployment rate for autistic people of working age at 18.2% - more than double the rate for people with disability generally, at 7.5%, and close to six times the rate for people without disability, at 3.1%. Labour force participation was 50.2%, against 84.9% for people without disability. ¹⁶ So about half of working-age autistic Australians are not in the labour market at all, and of those who are, nearly one in five cannot get work.

That pattern is not an artefact of one survey. Bury and colleagues followed 2,449 autistic adults across eight annual waves and found stable unemployment the single largest trajectory, at 1,189 people. ¹⁷ Baldwin, Costley and Warren surveyed 130 Australian adults and found people with the capacity and willingness to work facing significant labour market disadvantage and working below their qualifications. ¹⁸

I want to be plain about why an employment statistic belongs in an essay about loneliness. Work is where most adults get most of their social contact. It supplies the repeated, low-effort, structurally guaranteed exposure to other people out of which ordinary friendship forms. Remove it and you have not only removed an income. You have removed the mechanism.

The exclusion also registers physically. Trimmer and colleagues ran an ostracism task with autistic and non-autistic adults and found autistic participants rating their mood more negatively overall, and showing increased physiological arousal when excluded, which the control group did not. ¹⁹ Lin and colleagues found autistic traits associated with heightened sensitivity to rejection-induced social pain. ²⁰ Levi and colleagues, interviewing autistic adults about social pain, titled the paper with a participant’s phrase: feeling unwanted when nobody wants you around. ²¹

So: judged in seconds, mismatched in company, victimised at close to fifty per cent, locked out of the labour market at six times the general rate, and hurt more by all of it than the people doing it would guess.

That is the first push. Nothing in it required the autistic person to do anything at all.

Part 3:
The retreat that keeps you alive

Here is the part that gets misread most often, including by autistic people themselves.

We also shut people out. On purpose. And most of the reasons for it are good ones.

Neville and colleagues asked autistic adults in the UK what time alone is for. It is not idleness. Participants described it as protection against overwhelm and recovery from it, in environments they had chosen, doing things that let them regulate. ²² The framing in that paper is worth borrowing: reacting, retreating, regulating, reconnecting. Reconnecting is on the list. The retreat is not the destination.

What is being recovered from is mostly documented elsewhere. Higgins and colleagues, working with autistic adults as experts by lived experience, defined autistic burnout as a severely debilitating condition preceded by fatigue from masking, from interpersonal interaction, from cognitive overload, and from unaccommodating sensory environments. ²³ Three of those four are things other people do to you, and the fourth is the room they do it in. Quadt’s finding runs in the same direction: the places where people gather - pubs, offices, parties, open-plan anything - are built to a specification that is painful to sit in, and sensory difference tracks loneliness directly. ⁶ The retreat is not a retreat from company. It is a retreat from the conditions company is currently sold under.

But there is a limit on how much the retreat can actually fix, and it is worth being exact about where that limit sits.

The first part of it is what burnout takes. It does not only take your energy; it takes the compensation. Raymaker and colleagues’ definition is internal resources exhausted beyond measure, with no clean-up crew, ²⁴ and the specific loss people describe is the inability to engage with the things they are best at and care most about. So the solitude still arrives, and the thing solitude was for does not. You get the empty room without the flow it was supposed to buy.

The second is money, and it is the one nobody solves. The standard advice is to reduce external stress, and most autistic adults in trouble have already reduced it as far as they can while still affording to live. There is a floor under this, and the floor is rent. Which is where the employment figures from Part 2 stop being an abstraction: reduced hours, declined promotions and quiet exits are ordinary responses to overload, and each one lowers the ceiling on everything else.

There is also a positive reason to be alone, which gets less attention because it is harder to make sound like suffering. Murray, Lesser and Lawson’s monotropism model describes autistic attention as concentrating processing resources on relatively few things at a time, relatively intensely. ²⁵ Presented as a difference rather than a deficit, which is how they framed it. Being alone is where that concentration is available. An interrupted monotropic mind is not merely inconvenienced; it has lost the thing it does best.

Then there is the third reason, which is the one that starts turning the wheel.

You leave before you are left.

If enough encounters have gone badly, the cheapest available strategy is to stop having them. Umagami’s review lists negative experiences and learned helplessness among the factors raising loneliness in autistic adults. ³ Pearson and Rose make the parallel argument about masking - that the “choice” to mask is largely an illusion when the alternative is being treated as lesser ²⁶ - and the same structure applies to withdrawal. Choosing not to go is a choice in the sense that jumping is a choice when the building is on fire.

So the retreat has three motives stacked on top of each other: recovery, concentration, and defence. The first two are healthy. The third is a wound wearing the clothes of a preference, and from the outside all three look identical.

That is the problem. Nobody watching can tell them apart. Frequently, neither can I.

Part 4:
The paradox, stated plainly

Which brings us to the sentence that is hardest to say out loud without sounding as though you are contradicting yourself.

I don’t like people, I don’t trust them - I actively want to be, and prefer to be, alone - but, I also don’t want to feel lonely.

Both halves are true at once. Muris and colleagues named this the solitude-loneliness paradox: elevated loneliness co-existing with a stronger preference for solitude than non-autistic peers show. ⁴ It is not a contradiction and it is not indecision. It is two separate needs, both real, running at different frequencies.

The resolution is that solitude and connection are not opposite ends of one dial. Solitude is a recovery process with a duration. Connection is a requirement with a threshold. You can need eight hours of one and ninety minutes of the other in the same week, and be failing to get either.

Where it goes wrong is that the two get traded against each other by everyone involved. Other people trade them on your behalf - he prefers his own company, and the invitation stops coming. And you trade them yourself, because the recovery need is loud and immediate while the belonging need is quiet and slow. Declining tonight costs nothing tonight. It costs something in four years, and the bill does not itemise.

There is one more thing worth separating out. Being alone is restorative when it is chosen and corrosive when it is imposed. The same eight hours in the same room mean two entirely different things depending on whether anybody asked. And after long enough, you lose the ability to tell which one you are in, because there is no longer an invitation to decline.

The resolution nobody accepts: just be there

There is an obvious way to satisfy both needs at once, and most autistic people find it without being taught. You are in the room together and you are not interacting. Same sofa, different books. Same table, different screens. Company without conversation.

It is a real form of connection and not a workaround. Pritchard-Rowe and colleagues asked autistic adults about play across their lives and found solitary play serving a recuperative function, alongside a stated preference for parallel play and for playing with other autistic people. ²⁷ Presence without demand. It costs almost nothing to run, and it meets the belonging requirement without breaching the recovery one.

The trouble is that it is a two-person arrangement, and the other person has to agree to it.

To somebody who does not know what they are looking at, sitting in silence with a person you supposedly care about is not company. It is being ignored. Being ignored by someone in the same room reads worse than being ignored from a distance, because the capacity to engage is visibly present and visibly not being spent on them. So the offer of the highest-value thing an autistic person has - I want you here, and I want nothing from you - arrives as the lowest-value thing there is, which is indifference.

Heasman and Gillespie mapped this from both sides. Working with 22 pairs of people with Asperger’s syndrome and their family members, they found that family members systematically over-estimated the extent to which their autistic relative was anchored in their own perspective - that is, they read them as more self-absorbed than they were. ²⁸ The autistic participants, meanwhile, could accurately predict the lower scores they were being given, while disagreeing with them.

Sit with that second half, because it is the cruel part. You can tell you are being read as cold. You know the reading is wrong. And you have no way to correct it from inside the very behaviour being misread, because the correction would require the interaction whose absence is the problem.

What happens next takes one of two routes, and both of them end in the same place.

The first is that you let it happen. The other person concludes they are not wanted, and does the reasonable thing with that conclusion, which is to stop offering. You have then lost not just the friendship but the one mode of it that actually worked - the cheap, sustainable, non-depleting version you could have kept up for thirty years.

The second is that you see it coming and prevent it. You notice the reading being formed, you know where it goes, and you head it off by supplying the engagement they are looking for. You talk. You keep the eye contact up. You are present in the way the word is normally meant.

This is the route that looks like the good outcome, but it is the more expensive of the two.

What you have done is replace the cheap mode with the costly one, in the exact setting the cheap mode existed to protect. Scheeren and colleagues sampled 87 autistic adults through ordinary days and found masking rising when other people were present, rising further when those others were non-autistic - and higher masking at a given moment going with higher perceived stress at that same moment. ²⁹ Not remembered stress afterwards. Concurrent stress, measured while it was happening. So the room you were using to recover in is now a room you are working in.

And it does not stay in that room. Cage and Troxell-Whitman asked 262 autistic adults about camouflaging across contexts and found that both camouflaging heavily everywhere and switching between camouflaging in some places and not others were associated with poorer mental health. ³⁰ Having one relationship where the mask came off was doing work. Losing it is not a small change to the total.

Then there is what you actually bought. Participants in Pryke-Hobbes and colleagues’ study described friendships formed behind the mask as friendships with the persona rather than the person - reporting that nobody really knew them, that they had worked alongside the same people for years without becoming close to any of them. ³¹ So the manoeuvre succeeds on its own terms and fails at the thing it was for. They stay, and what they stay for is the performance, and you know it while it is happening.

The second route also has a shelf life. Nobody runs surface acting indefinitely in the place they live, and when it gives out it does not taper politely; it collapses into burnout, which is where Part 3 came in. At that point the withdrawal is not a preference or a strategy. It is a shutdown, and it arrives without warning, after years of apparent normality - which makes it look, from the other side, exactly like the coldness they suspected all along.

That is the wheel turning again, and this time nobody did anything wrong at all. One person offered connection in the only currency they had. The other person did not recognise it as value. And the accommodation - paying in a currency you do not have enough of - buys time rather than a solution.

That is the hinge. What follows is what happens when the wheel keeps turning.

Part 5:
The wider circle, and how it thins

Part 4 was about the person already in the room with you. This part is about everybody else - the wider circle, the ones whose entire relationship with you consists of asking whether you are coming.

For them the retreat is not misread as coldness. It is not read at all. It is simply a series of declines.

You say no to an invitation because you are three days into recovering from the last one. The person who invited you does not see recovery. They see a no. Repeat that four or five times and they stop asking, and they are not being cruel in stopping - that is the ordinary rule everyone runs. A person who says no five times is telling you something.

Except the message they received is not the message that was sent. Jaswal and Akhtar’s point applies exactly here: the behaviour that reads as absence of social interest has a different cause, and the cause is invisible. ⁷

The obvious remedy is to go anyway, and it is worth knowing what going anyway actually buys, because it is less than everyone assumes. Belcher and colleagues recorded eighty autistic and non-autistic adults in conversation and had ten-second clips rated by 127 non-autistic observers. Autistic participants were rated more poorly - and camouflaging intent did not predict first impressions at all. ³² That is a finding about strangers rather than about the people who already know you, and it is the harsher setting of the two: with a stranger there is no accumulated goodwill to draw on, and the judgement is made before you have said anything worth judging. Attending costs you the evening and does not reliably improve the impression.

I have written in my masking essay about why. The task in a new room is not performing a social convention; it is locating the convention, in a room where nobody has written it down.

So the wider circle contracts either way, and it contracts quietly. Nobody announces that they have stopped inviting you.

And now the loop closes for the first time, because that contraction changes you.

Cacioppo and Hawkley’s review of perceived social isolation is the clearest account of what happens next. Loneliness, they found, is associated with heightened sensitivity to social threat, and with a confirmatory bias in social cognition that is self-protective and, in their own phrase, paradoxically self-defeating. ³³ The lonelier you are, the harder you watch for signs of rejection; the harder you watch, the more you find; the more you find, the more you withdraw. The mechanism that evolved to get you back into the group is the mechanism that keeps you out of it.

You might hope that repetition dulls it. Büttner and colleagues tested exactly that, first with repeated exclusion in the lab and then with people who report chronic real-world exclusion, and found no habituation: exclusion hurt to the same extent regardless of how much of it a person had already had. What did change was the response to being included - repeatedly excluded participants became hypersensitive to inclusion. ³⁴ Hold on to that second result. It comes back in Part 7 and it does most of the damage.

Reinhard and colleagues reviewed experimental ostracism work across psychiatric conditions and describe the same shape as a vicious circle running between social exclusion and psychopathology. ³⁵ Nothing in that literature is specific to autism. It is what happens to human beings who get excluded often. Autistic adults are simply getting the standard human response to an above-standard dose.

This is the point at which the snowball has picked up its first real weight. The person is now not only excluded; they are primed to expect exclusion, and acting on the expectation.

Part 6:
The deprivation

There is a Facebook post that has been going around, and it is what made me want to write this essay. ³⁶

People don’t realize what emotional deprivation does to a person. When you’ve gone too long without love, affection, genuine care, or even a simple hug, loneliness starts changing the way you think. You become attached too quickly. You cling to people who were never meant to stay. You mistake the bare minimum for something extraordinary because you’ve been starving for kindness. You start giving endless effort just to receive a fraction of what you give back. You ignore red flags, overextend yourself, and slowly lose your self-respect trying to prove you’re worth loving. Your pride fades, your boundaries disappear, and you begin accepting things you once promised yourself you never would. Loneliness doesn’t just make you feel empty. Sometimes, it convinces you to settle for less than you deserve, simply because being treated poorly feels less painful than being alone.

I found it on a page called Suffering is my Hobby, and I do not know whether that is where it was written or only where I happened to see it. ³⁶ It is not research and it does not claim to be. But a lot of autistic people read it and recognised themselves, and that is worth taking seriously rather than dismissing, so I went looking for what the evidence does and does not support.

It supports more of it than I expected.

Start with the part about accepting things you promised yourself you would not. Griffiths and colleagues built the Vulnerability Experiences Quotient and administered it to autistic and non-autistic adults, finding elevated rates of adverse interpersonal experience in the autistic group and associations with mental health and life satisfaction. ³⁷ The vulnerability is measured, not asserted.

Pearson, Rees and Forster went closer in, asking 43 autistic adults about being taken advantage of by people they knew. Two things came out of it. People described repeated victimisation across different relationships - cycles rather than incidents. And they described difficulty recognising it as victimisation at all, and difficulty holding a boundary once they had, because of problems with trust, with compliance, and with reading what the other person actually intended. ³⁸ The title of the paper is a participant’s line: this was just how this friendship worked.

The companion paper is worse, and I am going to quote its title rather than paraphrase it, because paraphrasing softens it: I felt like I deserved it because I was autistic. ³⁹

So a person who has been short of connection for a long time, and who has been taught by repetition both that connection is scarce and that they are the defective party, is not well placed to walk away from a bad one.

Then the part about clinging. Van Asselt and colleagues interviewed 19 autistic adults who described heightened rejection sensitivity, and the accounts are hard reading: responses described as profoundly overwhelming and exhausting, accompanied by physical tension and pain, and by reliving past rejections in the present moment. Participants attributed it to a combination of autistic traits and a lifetime of actual rejection. ⁴⁰ Not imagined rejection. Actual.

The general model behind that is Downey and Feldman’s, and it contains the sentence that matters most in this essay. People high in rejection sensitivity anxiously expect rejection, perceive it readily in ambiguous behaviour, and react to it in ways that provoke the rejection they fear. ⁴¹ Gao and colleagues’ meta-analysis confirms the association with poor mental health outcomes across studies. ⁴²

That is the snowball written out as a mechanism, by researchers who were not thinking about autism at all.

And then the intensity of it. McDonald and colleagues pooled 35 studies and found a medium-magnitude difference in emotion dysregulation between autistic and non-autistic groups. ⁴³ Conner and colleagues, using a norm-referenced measure with clinical cut-offs, found community and inpatient autistic samples four and seven times more likely respectively to exceed the clinical threshold for emotional reactivity than a nationally representative comparison. ⁴⁴ Beck and colleagues asked autistic adults themselves what drives it, and got an answer that belongs in the title of the paper: the world is really not set up for the neurodivergent person. ⁴⁵

So the picture the post describes - deprivation, then attachment that is too fast and too hard, then loss of the boundary you thought you had - is not sentimental. It is the predictable interaction of measured rejection sensitivity, measured emotion dysregulation, and a genuine, documented shortage of supply.

Where I think the post overreaches is the moral framing at the end - settling for less than you deserve, as though the person made a bad trade through weakness. That is not what the evidence describes. It describes someone accurately reading a market in which their options are few, and then being blamed for the price.

The need you cannot meet on your own

There is one more layer, and it is the quietest one. It shows up as a particular kind of unease that I have felt myself and have seen other autistic adults describe: you trace the loneliness back, work out that what is missing is recognition from other people - being thought well of, mattering to somebody - and the discovery lands badly rather than usefully.

That reaction is worth examining, because the obvious reading of it is wrong. It is not embarrassment at wanting to be admired. It is the recognition of what kind of need has just been located.

Maslow’s original 1943 statement of the hierarchy included, alongside belonging, a need for “a stable, firmly based, (usually) high evaluation of themselves, for self-respect, or self-esteem, and for the esteem of others”. ⁴⁶ The esteem of others is in the specification. It is not an add-on for the insecure.

The hierarchy has taken a lot of deserved criticism since, mostly for the ladder - the claim that you must satisfy one level before the next comes into play. Tay and Diener tested the underlying needs across 123 countries and found that the needs themselves were largely universal and did predict wellbeing, but that they did not have to be met in order. Notably, positive feelings tracked social and respect needs specifically. ⁴⁷ So the pyramid is wrong about the sequence and right about the contents.

Which leaves the problem that has actually been identified. Ryan and Deci’s self-determination theory names three basic psychological needs: autonomy, competence and relatedness. ⁴⁸ Look at that list from the position of a capable, analytical adult in burnout. Autonomy you can arrange. Competence you can work at, and people in this position usually have. Relatedness is the only one on the list that another person has to agree to.

That is why it sucks. Not because needing recognition is shameful, but because he has correctly worked out that the remaining deficit is in the one category he cannot close by effort, intelligence or discipline - the three things that have solved everything else. It is the difference between a hard problem and a problem that is not yours alone to solve.

And for an autistic adult that is not a general truth about the human condition. It is specific, because Part 2 was about how few people are available to agree to it.

This is where the loop takes its quietest turn. A need that goes unmet for long enough stops being experienced as a need and starts being experienced as a fault - not I am not getting this, but there is something about me that does not get this. Van Asselt and colleagues tested that shape in 831 autistic adults on the Netherlands Autism Register, examining internalised stigma and loneliness as pathways between minority stress and both stress and life satisfaction. ⁴⁹ Internalised stigma is the technical name for having taken the outside view of yourself and adopted it as your own.

Which is the worst thing the snowball does, because it is the point at which it no longer needs anybody else to keep it rolling.

So: excluded, withdrawn, deprived, and now holding a need that cannot be met alone and has started to look like a personal defect. That is four turns, and the person has still done nothing wrong.

Part 7:
All or nothing, and why it ends the thing it wanted

Now the turn that does the most damage, and the one I have the least clean evidence for. ⁵⁰

When somebody does arrive - a friend, a partner, anyone who seems to actually see you - the response is not proportionate. It is total. Everything goes into it. Every spare hour, every message answered inside a minute, the whole of your attention pointed at one person like a lamp.

I do not think this is a character flaw and I do not think it is desperation, though it can look like both. Two things are stacked underneath it.

The first is the ordinary autistic attentional style applied to a human being. Monotropism describes a mind that concentrates its resources on few things, intensely. ²⁵ Nothing in that model exempts people from being one of the things.

That is the point I want to make carefully, because it is the one people flinch at. The special interest is normally discussed as a topic - trains, taxonomies, a discography, a decade of military history. Grove and colleagues surveyed autistic adults on their interests and found them broader than the stereotype allows, and found motivation to pursue them associated with higher subjective wellbeing and better life satisfaction, including satisfaction with social contact. ⁵¹ The special interest is not a symptom. It is one of the few reliably good things in the literature.

Now consider what happens when the subject of that attention is a person.

Everything that makes a special interest rewarding still applies. The immersion. The wish to know everything. The pleasure of the detail nobody else noticed. The difficulty stopping. But a topic does not have a view about being studied, and a person does. Trains do not notice the intensity. People notice immediately.

What the research does and does not cover

I went looking for work on this and found something worth reporting in itself. ⁵²

The literature on special interests is a literature about topics, and I could not find a peer-reviewed taxonomy that treats a specific living person as a category of interest. The one place where science has looked closely at autistic people directing sustained, intense attention at another person is the forensic literature - which means the only well-documented version of this is the version that ended badly enough to involve the police.

I want to be exact about what that does and does not tell us, because the risk of being misread here is high. It is a statement about where the research lamp has been pointed, not about where the behaviour usually leads. Nothing about intense attachment predicts a criminal outcome, and I am not suggesting it does. The problem is that when the only available evidence comes from the cases that ended in court, the evidence itself is skewed - and you have to read it knowing that.

Stokes, Newton and Kaur compared parental reports for 25 autistic and 38 non-autistic adolescents and adults. The autistic group were more likely to focus their attention on celebrities, strangers, colleagues and ex-partners, and to pursue the target for longer, and were more likely to engage in courting behaviour rated inappropriate. ⁵³ Murphy and colleagues note that stalking is among the behaviours that bring autistic people into contact with the criminal justice system, and that treatment models adapted for this population barely exist. ⁵⁴

Three cautions, and I mean all of them.

The first is the sample: 25 people, by parent report rather than self-report, in 2007. That date matters more than it looks. The study predates the DSM-5 by six years, so its participants were diagnosed under DSM-IV, when autistic disorder, Asperger’s disorder and PDD-NOS were still three separate categories rather than one spectrum. ⁵⁵ The group it describes is therefore not the group the word autistic picks out today, and parent report is not the same instrument as asking the person.

The second is the inference. Nothing in it says autistic people are dangerous. The great majority of what this section describes never goes anywhere near a criminal threshold; it ends in an unanswered message and a friendship that quietly stops.

The third is the shape of the evidence base, which I have already flagged and will repeat because it is the one most easily forgotten. A literature that only studies a behaviour at its most harmful will make that behaviour look worse than it is, every time.

But the Stokes paper contains one finding that explains far more than the headline does, and it is not about pursuit at all. The autistic participants were reported as having learned romantic skills from essentially nowhere: not from parents, not siblings, not observation, not media, not sex education, not peers. ⁵³ Every channel the rest of the population uses to absorb the rules had failed to deliver.

So what is being described is not excessive desire. It is undiluted attention aimed at a person, by somebody who was never given the manual - and who therefore does not know that there is a rate at which this is supposed to be done.

Which is exactly where the rest of this part goes.

The second thing stacked underneath it is Büttner’s result, which is the one I did not expect and cannot unsee. Repeated exclusion does not blunt you to rejection; it makes you hypersensitive to acceptance. ³⁴ Which means the reaction is not out of proportion to the person. It is in proportion to the drought.

There is empirical shape to it too, though I have to be careful about what shape exactly. Sedgewick, Hill and Pellicano compared the friendships of autistic and non-autistic adolescents and found autistic girls’ friendships more intense, with an individual best friend often becoming the sole focus of their social lives, alongside more conflict and more difficulty resolving it. ⁵⁶ Fewer connections, each carrying more load.

Now the honest caveat, because I have just done something the reader is entitled to object to. That study measured the structure of a social network - how many friends, how exclusive, how much conflict, how well repaired. It did not measure attention, and it did not describe anybody’s best friend as a special interest. Two paragraphs ago I said there is no research treating a person as the object of an autistic special interest, and that is still true; a finding about exclusive friendships is not the same claim in different clothes. What I am doing is putting a structural finding next to an attentional model and observing that they point the same way. That is an inference, it is mine, and it is not what Sedgewick measured. ⁵⁰

I also have to retract the line I would rather have written here, which is that none of this stops at eighteen. Sedgewick and colleagues went on to study adult women and found some of the pattern persisting - no network of friends to supply relationship advice, a tendency to take sole blame for conflict or to write the friendship off and withdraw, and high rates of victimisation. But they also found the autistic women happier and more self-assured in their adult relationships than they remembered being as adolescents. ⁵⁷ So it does change. Whatever the adolescent picture is, it is not a life sentence, and I am not going to imply it is because the darker version reads better.

One further limit on both of those studies. They sampled girls and women, because that is where this particular research line went, and I have no evidence that the pattern is confined to them - nor any that it is not. I am describing the shape of an available literature, not a difference between groups.

And it is made worse, not better, by scarcity. If you have one friend, that friend is not one of your friends; that friend is the entire category. The stakes on every interaction go up accordingly, and so does the vigilance, and rejection sensitivity is already running hot. ⁴⁰

Here is where it breaks, and the reason is not the one people assume.

The obvious explanation is that you gave too much and frightened them off. The evidence does not really support that as stated. Collins and Miller’s meta-analysis of self-disclosure found that people who disclose intimately are generally liked more, not less. ⁵⁸ Openness is not the problem. Intensity is not, by itself, the problem.

What their third finding points at is the actual variable: people come to like others as a result of having disclosed to them. The effect runs on reciprocity. A relationship is a two-way ratchet, and what governs it is the matching - each side moving at roughly the pace of the other.

Which is the part that does not get said plainly enough. Relationships are not willed into existence, and they are not owed to anybody for being decent. They take a time investment, and trust is earned incrementally, in instalments whose size the other person sets. That is the schedule. And the autistic error is not refusing to pay it. It is trying to pay the whole thing on the first day, because nobody mentioned it was an instalment plan.

Which means the failure is not caring too much. It is misreading the pace. And misreading the pace is precisely the disability. The rate at which intimacy is supposed to increase in a new friendship is an unwritten local convention, learned by observation, and never once stated aloud. It is the same problem as every other social convention: perfectly executable once you know it, and close to undetectable if you have to infer it.

So the giving is not wrong in kind. It is wrong in timing, against a schedule nobody published. And a gift given at the wrong point in the schedule does not read as generosity. It reads as pressure.

Then the other person pulls back. Sometimes gently, sometimes not. And every part of the machinery described above is now pointed at that retreat: heightened threat detection ³³, reliving of past rejections ⁴⁰, the reaction that provokes the thing it fears ⁴¹, and an emotional response substantially more likely to exceed the clinical threshold for reactivity than anyone else’s ⁴⁴. You reach harder, exactly when reaching harder is the worst available move.

The relationship ends. And it ends in a way that confirms every prediction the loneliness made.

That is the snowball completing a full turn. You are now lonelier than you were, more certain that closeness ends badly, more primed to detect the ending early, and another friend short.

Part 8:
A warning about this argument

I have to stop here and say something about the shape of what I have just written, because it has a dangerous twin.

Everything above says: this is being done to autistic people, the mechanism is not their fault, and the explanation lies outside them. That is where the evidence points. It is also, word for word, the opening move of every grievance movement that has ever recruited a lonely person - a real pain, correctly identified, followed by a group of people handed over to carry the blame for it.

You can watch the deal being offered on any video platform. Somebody posts an honest account of a real and widespread form of loneliness; the recommendation engine then serves up a response explaining which category of people is responsible for it. The second video is more satisfying than the first, which is precisely the problem, and it is why the recommendation was made.

I am not going to name the particular quarrels this happens in, because the specifics are a distraction and rehearsing them only circulates them further. The structure is the point, and the structure is portable. It attaches to any diffuse, structural, hard-to-fix harm, and it works best on people isolated enough to be grateful for an explanation - which, if the rest of this essay is right, describes a great many autistic adults.

So here is the version of it built for us.

If Part 2 is right - if autistic people are judged in seconds, mismatched in company, victimised at close to half, and locked out of work at six times the general rate - then it is a very short step to concluding that non-autistic people are the problem. That step is available. It is emotionally satisfying. And a certain amount of autistic community writing has taken it.

It does not survive this essay’s own evidence.

Milton’s double empathy argument is mutual by construction: two groups failing to read each other, not one group failing the other. ¹⁰ Crompton’s finding is that information transfer degrades in mixed pairs - not that non-autistic communication is defective. ¹¹ Heasman and Gillespie found the misreading running in both directions at once, with each side misjudging the other. ²⁸ And the Sasson result, which is the most damning single finding in Part 2, has a companion I have already cited: the same kind of observers form better impressions once they know something about autism. ⁵⁹ They were not malicious. They were uninformed, and informing them worked.

That is a considerably less satisfying story than one with villains in it. It is also the one the data supports.

Then the harder half, which I would rather not have to say to people who are lonely.

The mechanism is not your fault and somebody therefore owes me are different sentences, and the second does not follow from the first. Nobody is obliged to be close to me. A person who declines an invitation is not malfunctioning and is not the cause of the snowball. The exclusion in Part 2 is a population-level pattern produced by mismatched communication, inaccessible environments and stigma - and Botha and Frost are precise that minority stress operates through social position rather than through anybody’s malice. ⁶⁰ It is not a debt owed by the last individual who said no.

I want to be plain about why this section is here at all. An essay that explains your loneliness as something done to you is a comfort, and a comfort that load-bearing is worth a great deal to somebody who has been without one. That is exactly what makes it worth hijacking. Anyone offering you a category of people to blame for this is offering you a worse deal than the one already on the table, because the category is not the mechanism, and resenting them repairs none of it.

This essay is a description of a mechanism. It is not a grievance, and it is not an invoice.

Part 9:
What the loop costs, and where it can be broken

The costs are not abstract, and I am not going to soften them.

Holt-Lunstad and colleagues’ meta-analysis puts social isolation at a 29% increased likelihood of mortality, loneliness at 26%, and living alone at 32% - effects comparable to well-established risk factors. ⁶¹ Pelton and colleagues tested the interpersonal theory of suicide in autistic and non-autistic samples and found thwarted belongingness and perceived burdensomeness operating in both, at higher levels in the autistic one. ⁶² Cassidy and colleagues connected camouflaging to thwarted belongingness and to lifetime suicidality. ⁶³

Botha and Frost’s minority stress account is the frame that holds all of it: the harm follows social position, stigma and concealment rather than the neurology. ⁶⁰ Slavich’s social safety theory makes the biological version of the same point, treating the perception of social safety and threat as a central regulator of health. ⁶⁴

So the loop is not a mood. It is a health outcome.

Where can it be interrupted? On the evidence, in four places, and only one of them is inside the autistic person.

Stop trying to be received well and start being known. 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. ⁵⁹ It is one of very few things in this literature with a measured effect on how autistic people are received, and it works by changing the observer rather than the observed.

Change the pairing. The double empathy work is not a consolation prize. Information transfer between autistic people is as good as between non-autistic people ¹¹, and real-world interaction outcomes are better in matched pairs than mixed ones. ¹² Crompton and colleagues asked autistic adults directly about their relationships with autistic and non-autistic people, and the title is a participant’s line: I never realised everybody felt as happy as I do when I am around autistic people. ⁶⁵ Autistic community is not a lesser substitute for the real thing. On the measured outcomes, it is the real thing.

The thing being missed, when autistic adults describe what is absent from their lives, is usually not a friend. It is a collective - a set of people who are around, doing things, in a way you are part of without having to negotiate entry every time. Botha, Dibb and Frost investigated autistic community connectedness and found it made of three components: belongingness, social connectedness, and political connectedness. ⁶⁶ That is not a support group. It is a culture, and it is one you can be a member of.

It also does not have to be local, which matters more than it sounds. Most autistic adults I know found their people in a place with no postcode - a forum, a server, a group organised around a shared interest rather than a shared street. That gets dismissed as second-best community, and I do not accept that it is. A community assembled from the small number of people who fit, drawn from everywhere, and carried with you when you move, is not a substitute for the town you happen to live in. For a lot of us it is the only version that was ever going to work.

Change the room. Sensory environment is not a comfort issue; it is an access issue, and it is associated with loneliness directly. ⁶ A social space you cannot physically tolerate is a social space you are barred from, whatever the invitation says.

And, inside the person, one thing only: name the pace problem instead of trying to feel it. If the schedule of intimacy cannot be read by instinct, it can sometimes be handled explicitly - saying what you need, asking what they need, and treating the answer as information rather than verdict. That is not a cure, and I want to be careful not to present it as one. It is a workaround for a disability, run consciously, at cost, which is what most autistic accommodations are.

One thing is worth adding on what does not work, because it is the strategy most of us reach for first. Su, Schiltz and Lord looked at how autistic adults cope with loneliness and found that most reported temporary distraction - and that people who use distraction tend to be lonelier, not less. ⁶⁷ Filling the evening is not the same as filling the gap.

And one correction, because it matters

There is a reason autistic adults give for not pursuing the obvious help, and I have given it myself more than once: therapy does not really work on me, because I am an analytical thinker rather than an emotional one.

I want to take that seriously and then disagree with it, because every time I said it, it was wrong.

What is being described there has a name. Alexithymia is difficulty identifying and describing one’s own emotional states, and it is not rare in this population: Josyfon and colleagues assessed 190 adults attending a tertiary autism clinic and classified 66.3% of them as alexithymic. ⁶⁸ Two thirds. So “I am analytical rather than emotional” is, for a great many of us, not a personality type. It is reduced access to internal data that is still there and still operating - which is also why loneliness can accumulate for years without registering as anything, until it arrives all at once as burnout.

That matters for the conclusion drawn from it. Emotion-focused therapy assumes a degree of access to your own feelings that most autistic clinic attendees do not have, so its failing is a fit problem rather than a verdict on the client. Paynter, Sommer and Cook asked 130 autistic adults to rate 55 possible adaptations to therapy and found around half rated positively - with wide variation between individuals, which is itself the finding. ⁶⁹ What follows from that is not try harder at therapy. It is that the standard version is one configuration of a thing that has many, and one configuration not fitting is information about the configuration.

And it is not only the method that varies. It is the person delivering it, which I have written about in my unmasked at midlife essay and will repeat here because it belongs to this argument too.

The gap here is documented, and it is documented by the clinicians themselves. Cooper, Loades and Russell surveyed therapists on their skills, experience and confidence in working psychologically with autistic people, and found confidence varying considerably across skills, adaptations made unevenly, and challenges the therapists named themselves - concluding that well-designed training and supervision were required. ⁷⁰ Moore, Larkin and Foley’s synthesis of 13 studies of professionals adapting interventions for autistic adults found the same picture from the other end: adaptation treated as a highly individualised process, shaped as much by an individual practitioner’s own traits and experience as by anything systematic, and constrained by service-level factors. ⁷¹

Read that as a client rather than as a researcher and it says something uncomfortable but useful. Whether the therapy fits you depends substantially on which individual you happen to get, because the profession has not standardised this and the practitioners know it.

That is not a reason to avoid therapy. It is a reason to shop. The specific failure mode to watch for in this territory is a clinician who hears “I have almost no friends” and sets about correcting your social skills - when Part 5 has already established that effort at the presentation is not what determines how you are received. Asking what training and experience someone has with autistic adults before you commit is not rudeness. On this evidence it is the single most informative question available to you, and you are entitled to keep asking it until you get an answer you believe.

What I will not do is end this by telling anyone to put themselves out there more. The evidence does not support that instruction. Autistic adults are not failing to want connection ⁷, and trying harder at the presentation does not improve how it lands. ³² The people telling us to be more social are describing a mechanism that works for them and has been measured not working for us.

In closing:
A disability, filed as a preference

Being alone is fine. I am good at it, and I intend to keep doing it. What is not fine is that the choosing gradually stopped being a choice - that the calendar emptied for reasons that were mostly other people’s, and then the emptiness got explained back to me as my personality.

Nor is the problem solved by putting me in a room with people. The state this essay has mostly been about is not the empty room. It is the full one: standing in it, following exactly what everybody is doing and why, and having no way in.

And what is least fine is the last turn, the one nobody sees. That when somebody finally does turn up, having gone too long without makes you hold on too tight, too fast, at the wrong point on a schedule you were never shown - and that this is the most reliable way there is to make them leave.

The snowball is not that autistic people are lonely. It is that being lonely makes you worse at not being lonely.

There is one more thing to say, and I have kept it until last because everything above is a consequence of it.

Autism is a disability. Not a metaphor, not a style, not a way of being quirky - a disability, counted as one by the statistical and legal machinery of most English-speaking and European countries. ⁷² The employment figures in Part 2 came from a national survey called Disability, Ageing and Carers, and I let that pass without comment at the time. ¹⁶

Now look at what the disability is of.

The DSM-5 defines autism by persistent deficits in social communication and social interaction across multiple contexts, and requires all three of its sub-criteria - no partial credit. ⁵⁵ The ICD-11 puts it as persistent deficits in the ability to initiate and to sustain reciprocal social interaction and social communication. ⁷³

Initiate and sustain. That is not the description of a symptom. It is a description of what a relationship is made of, written by committees who were not thinking about loneliness at all.

Every other disability has an accommodation you can name. A ramp. A hearing loop. A screen reader. Flexible hours, a quiet room, a bigger font. The accommodation sits between the person and the barrier and does its work without anybody having to feel anything about it.

There is no ramp for a conversation. The barrier is other people, so the accommodation would have to be other people, and other people cannot be required to install one.

That is the hard version of everything above. The snowball is not an unfortunate side effect of being autistic. It is the disability doing exactly what the manuals say it does, in the one domain where the damage compounds - because each failed attempt at connection reduces the number of future attempts available. A missing ramp is missing again tomorrow, but it does not make the building harder to enter next year.

So when somebody says but you like being on your own, listen to what is actually being recorded. A disability is being observed in the only place it can be observed - in the space between people - and filed as a preference.

It is the cheapest sentence in the language. It ends the conversation, it costs the speaker nothing, and it costs the person it is said to another year.

⁂

---