DEEP READ · 942
The Legacy of Autism and the Future of Neurodiversity · Steve Silberman · 2015
The explosion in autism diagnoses over the past three decades is, for the most part, not an explosion in cases. It is what happens when you widen the ruler several times and keep reporting the readings as if the ruler never changed. The book proves this by telling a distorted disciplinary history: a diagnosis defined extremely narrowly in America in 1943 kept almost everyone out for forty years — and, along the way, assigned the blame to mothers. Only when someone widened it into a "spectrum" in the 1980s did the people who had been excluded pour into the statistics all at once, which looked like an epidemic. And the premise the book really wants to replace is the default one: autism is not a broken version of an ordinary brain, but a configuration that has always been in the human gene pool.
Steve Silberman was a science journalist who came out of Wired. The book appeared in 2015, won that year's Samuel Johnson Prize (now the Baillie Gifford Prize) and carries a foreword by Oliver Sacks. Its starting point was his 2001 Wired article "The Geek Syndrome," on the cluster of autism diagnoses in Silicon Valley engineering families. That piece left him with a question he could not answer: is this new, or has it always been here and only now become visible? The five hundred pages of this book are the belated answer. (Silberman died in 2024.)
Start with the numbers. The earliest epidemiological surveys in the 1960s — Victor Lotter's 1966 survey in Middlesex, England, among them — put autism at roughly four or five per ten thousand. Current CDC surveillance in the United States is in the range of one in 31 to one in 36. That looks like a hundredfold increase.
But look at what happened in between. The diagnostic manual was rewritten four times. DSM-III (1980) listed "infantile autism" as a separate diagnosis for the first time; DSM-III-R (1987) replaced it with a looser symptom checklist; DSM-IV (1994) added two capacious new pockets, Asperger's disorder and PDD-NOS; DSM-5 (2013) merged everything into one continuous autism spectrum disorder. Each rewrite converted some child from "not counted" into "counted."
The hardest evidence for this reading is called diagnostic substitution. The logic is simple: if the people in the autism column are genuinely new, the other columns should be unaffected; if they have merely been moved across from somewhere else, the other columns should shrink in step. Researchers went through decades of US special-education records and found the second pattern — the autism category rising while "mental retardation" and "learning disability" categories fell. The same children, filed in a different drawer.
One analogy settles the intuition in a second: when schools in the mid-twentieth century stopped forcing left-handed children to write with their right hands, the measured proportion of left-handers shot up and then levelled off. Nobody went looking for an environmental cause of left-handedness, because everyone understood the counting rule had changed, not the hands. Most of the autism curve is the same phenomenon.
An honest caveat: "most" is not "all." Broader criteria, better awareness and service-linked diagnosis explain the bulk of the rise, but whether they explain one hundred percent of it is still debated (rising parental age and improved survival of very preterm infants are among the factors under test). The book's position — that this is essentially a measurement problem — is a strong one, but it is not a closed case.
In 1943 Leo Kanner, a child psychiatrist at Johns Hopkins, published a description of eleven children under the title "Autistic Disturbances of Affective Contact" — the clinical naming of autism. His definition had three features, each of them consequential: very rare, beginning in early infancy, defined by severe impairment. Children who fit were autistic. Children who nearly fit — the ones who could speak, could attend school, had ordinary IQs — were excluded.
The cost was more than missed diagnoses. A diagnosis allocates not just a label but resources. Children kept outside the door never entered the research samples, so the research kept seeing only the most severe end, so the definition never moved: a closed loop. Meanwhile those children were filed under childhood schizophrenia, mental deficiency, emotional disturbance — and many of them were institutionalised.
This is the methodological lesson worth carrying out of the book: in medicine, "how many people have this?" can never be answered independently of "what have we defined it as?" And definitions are written by people. The rule is not specific to autism — prevalence curves for ADHD, depression, hypertension and prediabetes all deserve the same first question: has the threshold been moved lately?
Kanner's early writing noted that these children's parents were often highly educated and seemed emotionally reserved, and hinted that the parent-child relationship might be implicated. Bruno Bettelheim took the hint and pushed it to its limit: in The Empty Fortress (1967) he explained autism as a child's defensive withdrawal from a hostile mother, using an analogy that is suffocating to read now — comparing these mothers to concentration camp guards. The "refrigerator mother" became a verdict handed down on an entire generation of parents.
The mechanism matters more than the specific error: when a condition has no known cause and its presentation depends heavily on interaction, the person most likely to be named as the cause is whoever is closest to the child and has the least standing to object. Mothers were not indicted by evidence. They were indicted by proximity. (The sequel is instructive: Bettelheim's credentials were later found to be substantially fabricated, and allegations of abuse at the school he ran surfaced after his death.)
The demolition came from Bernard Rimland, a psychologist and the father of an autistic son. His 1964 book Infantile Autism argued systematically for a neurobiological basis rather than a parenting one, and he went on to build the first American parent organisations. It is one of the most striking turns in the story: the people who removed the causal blame from mothers were not the profession, but the accused themselves. Kanner eventually gave ground in public too, telling a parents' conference in 1969 something to the effect of: I hereby acquit you people as parents.
The book does not sanctify Rimland either, and it is better for it: the same man later promoted megavitamin regimens, chelation and other unevidenced interventions, and became deeply involved in the anti-vaccine movement. A person can be right and then wrong about the same subject, and this account does not hide it.
Almost simultaneously, the Viennese paediatrician Hans Asperger published, in 1944, his account of a group of boys under the name "autistic psychopathy." The decisive difference: he explicitly described a continuum, running from children needing lifelong care all the way to adults who were merely odd, intensely focused and socially awkward. He also emphasised their particular abilities — he called some of them "little professors" — and argued that their focus and pattern recognition were assets in the right job.
That road was buried for nearly forty years, partly for linguistic reasons (a German paper published in wartime), partly institutional ones. Silberman adds a sharper conjecture: Kanner probably knew about the Vienna work — two of Asperger's colleagues, Georg Frankl and Anni Weiss, emigrated to the United States and joined Kanner's team at Hopkins, and Frankl personally examined the child Kanner called Case 1. It is the book's most accusatory and most contested inference.
And one thing must be said here rather than saved for the end: the book portrays Asperger as a man who stressed his patients' value in order to shield them from the Nazi "euthanasia" programme, and that portrait has not survived 2018. Archival research by the historians Herwig Czech and Edith Sheffer showed substantive cooperation between Asperger and the Nazi racial-hygiene apparatus, including referrals of children to Am Spiegelgrund, where children were systematically killed. Read that chapter as a narrative that has since been overturned.
| Kanner (1943, Baltimore) | Asperger (1944, Vienna) | |
| Scope | Very narrow: a rare, severe condition | A continuum, from severe to merely "odd" |
| Language | Mostly described as severely impaired | Included fluent, precociously verbal children |
| Special abilities | Part of the symptom picture | Usable strengths (the "little professors") |
| Implied cause | Early hints at parenting, later inflated into "refrigerator mothers" | Constitutional / hereditary |
| Legacy | Dominated the Anglophone world for forty years | Buried until Lorna Wing revived it in 1981 |
Two first drafts of the same phenomenon. The narrow one won — hence forty years of missed children and forty years of maternal guilt.
The turning point came from the British psychiatrist Lorna Wing — also the mother of an autistic daughter. With Judith Gould she ran the Camberwell study in the late 1970s: instead of screening against Kanner's checklist, they went through one London borough child by child, looking at everyone with social impairments. The result: very few children met Kanner's strict criteria, but several times as many had the same core difficulties — in social interaction, communication, and imagination with rigid repetitive behaviour, what she called the triad. Those children had always been there. They just did not match the form.
In 1981 she reintroduced Asperger's 1944 paper to the English-speaking world, coined the name "Asperger's syndrome," and pushed the term autism spectrum into general use. Much of the subsequent climb in prevalence begins right there: not a change in incidence, a change in mesh size.
While we are here, a near-universal misconception worth correcting: the spectrum is not a line running from mild to severe. It is closer to a multi-dimensional radar chart. One person can be highly verbal and catastrophically overloaded by sensory input; another can be nonspeaking and intensely sociable. "Mild versus severe" is actively harmful, because it denies support to people who "seem fine" and denies an inner life to people who need a great deal of support.
In 1998 Andrew Wakefield published a paper on twelve children in The Lancet suggesting a link between the MMR vaccine and autism. It was fully retracted in 2010, the year he was struck off the UK medical register, and epidemiological studies with samples in the millions have repeatedly found no association. But the panic had done its work: vaccination rates fell in several countries and measles came back.
The book's contribution is not one more denunciation of Wakefield; it is the explanation of why the match caught. Three conditions had to hold at once:
Together those three make a general template worth keeping: a widening case definition, plus an onset window that happens to coincide with some near-universal event, plus authorities with no explanation, equals an attribution disaster. When you see that structure on some other issue, treat it as an alarm.
Neurodiversity dates from the late 1990s: the Australian sociologist Judy Singer used and popularised it in her academic work, and the journalist Harvey Blume brought it to a wide audience in The Atlantic in 1998. The analogy is biodiversity: an ecosystem's robustness comes from genetic variety, not from every organism being identical. By extension, autism, ADHD and dyslexia are points on the natural variation of human neurology, not defective units of one standard machine.
The blunter version: this is not a computer running a broken copy of Windows; it is a computer running a different operating system. That it won't run software written for the other one does not make it faulty. (The limits of the analogy are in the criticism section below.)
Institutionally this is a front in the fight between the medical model and the social model of disability. The medical model asks what this person lacks and how to fix it. The social model asks what the environment lacks that stops this person functioning. Is a wheelchair user kept out of the building by their legs, or by the stairs? Neurodiversity's claim is that a great deal of autistic disability is stairs: open-plan noise and lighting, interviews conducted entirely in speech, social protocols that demand eye contact. Replacing the stairs is enormously cheaper and more feasible than replacing the legs.
The sharpest statement of the position comes from autistic people themselves. The sense of Jim Sinclair's 1993 address to parents, "Don't Mourn For Us," is this: when you say "I wish my child did not have autism," what you are actually saying is "I wish the child I have did not exist and a different child were here instead" — because autism is not a shell around the child, it runs through everything about how he perceives and thinks. It is a harsh sentence, and it identifies precisely what is logically awkward about "cure" as a goal.
The "tribes" of the title are not a figure of speech. Autistic people emerged as a self-aware group immediately after one particular technology arrived: text-based, asynchronous networked communication. The mechanism deserves unpacking.
Face-to-face conversation is close to the worst possible medium for autistic people: real-time processing of expression and tone, instant response required, heavy sensory noise, and an expectation of eye contact throughout. Mailing lists, forums and later social networks removed nearly all of that: you can read slowly, write slowly, skip the face, and leave whenever you need to. The medium happened to rebuild sociality around the way autistic people actually process. So in the early 1990s came the first autistic mailing lists, Autism Network International (founded by Jim Sinclair and others), and Autreat — an annual gathering run by autistic people and designed around their own sensory needs.
The general insight: whether a group can form often depends less on how many of them there are than on whether a medium exists that its members can actually use. Autistic people have always existed, but before asynchronous text they were scattered beyond each other's reach — so they were isolated eccentrics rather than a group that knew it existed. What produced the tribe was not a discovery. It was bandwidth.
Bringing neurodiversity back down to practice, the real argument is about the objective function of intervention. Twentieth-century behavioural treatment — exemplified by Ivar Lovaas at UCLA — stated its target plainly: make the child "indistinguishable from peers." His 1987 study claimed a substantial proportion of intensively treated children reached "normal functioning," and that lineage (applied behaviour analysis, ABA) became standard in many countries.
The book does not skip the provenance: early Lovaas programmes used aversives, including slaps and electric shock, and in the same period he was involved in research aimed at "correcting" effeminate boys. Put those together and the nature of the objective function is clear — it was not optimising for whether the child was doing well, but for whether he looked normal.
That is the substitution the book is really arguing for: replace "indistinguishable" with "functioning." The first is scored by observers; the second by whether the person can communicate, manage daily life, and suffer less. They often coincide — but when they conflict, as when hours of training are spent extinguishing a harmless self-soothing movement like rocking or hand-flapping, the first goal extracts a real cost and buys only someone else's comfort. (In fairness: contemporary ABA has largely abandoned aversives and the field has been correcting itself, but debate over its goals and long-term outcomes is far from settled.)
For a five-hundred-page book the structure is a single line — a causal trace that uses history as its evidence: here is an anomaly, here is where it came from layer by layer, here is the premise you should therefore drop.
| When | What happened | Effect on "how many" |
| 1943 / 1944 | Kanner's narrow definition vs Asperger's continuum | The narrow one wins; almost nobody counts |
| 1950s–60s | "Refrigerator mother" spreads; Bettelheim amplifies it | Causation misassigned to parents; research stalls |
| 1964 | Rimland argues for a neurobiological basis; parents organise | Causation shifts; the door stays narrow |
| 1979–1981 | Wing & Gould's Camberwell study; "spectrum" and Asperger's syndrome | The door widens — the curve starts here |
| 1980–2013 | Four DSM rewrites, each looser than the last | Diagnostic substitution: people move across columns |
| 1988 / 1998 | Rain Man creates public awareness; Wakefield lights the vaccine panic | Referrals rise again; attribution goes badly wrong |
| 1990s– | The internet and ANI; the word "neurodiversity" appears | The goal shifts from cure to fit |
Every upward step in the curve stands next to a rewritten definition, not a new cause.
A word on Rain Man: the 1988 film almost single-handedly made "autism" a household word, and it also installed a stubborn stereotype — one of the character's real-life models, Kim Peek, was not autistic (he had agenesis of the corpus callosum), and the great majority of autistic people have no extraordinary memory or calculating ability. A film can open a door and post a misleading sign beside it.
The most important, already flagged above: the book's rehabilitation of Asperger must now be treated as overturned. Herwig Czech's archival research in Molecular Autism (2018) and Edith Sheffer's Asperger's Children (2018) together documented substantive cooperation with the Nazi eugenics apparatus, including referrals of children to institutions such as Am Spiegelgrund. The portrait of a protector, and the reading of "little professors" as a shielding strategy, does not match the archives. (Details remain debated, and some scholars argue parts of the case need more cautious handling, but the overall reassessment has happened.) The right way to read the book is: accept the structural argument, and strike through the Asperger chapters.
Criticism 2: Kanner is cast as the villain, and the script is too tidy. The inference that he knew of the Vienna work and concealed it rests on circumstantial personnel evidence — Frankl and Weiss really did go from Asperger's clinic to Kanner's team — but nothing documents deliberate suppression. The book also underweights his contribution: he was the first to pick these children out of the "childhood schizophrenia" catch-all, and he publicly cleared parents of blame in 1969. The hero-and-villain frame makes the book readable and makes its historical judgements too hard-edged.
Criticism 3: one end of the spectrum gets very little air. This is the sharpest fault line inside the autism community. The book's centre of gravity — and its "geek syndrome" starting point — sits with people who are verbally fluent, employable and able to advocate for themselves. People who are nonspeaking for life, with significant intellectual disability and self-injury, needing round-the-clock care, along with their ageing parents, often find no place to stand in the neurodiversity narrative. The 2021 Lancet Commission proposed the term "profound autism" precisely so that this group would not be averaged away. Honestly stated: reframing autism as difference rather than deficit is liberation for some people and, for some families, sounds like a denial of what they deal with every day. Both can be true at once.
Criticism 4: the "different operating system" analogy has a limit. It demolishes the "broken machine" premise beautifully, but pushed all the way it slides into "nothing here needs medical attention" — and comorbid epilepsy, severe sensory pain, self-injury, sleep and gastrointestinal problems are real medical problems that no amount of environmental design will solve. The defensible version of neurodiversity is not "treat nothing." It is "distinguish: which suffering comes from the person, and which from an environment that refuses to accommodate them" — the first deserves treatment, the second means the environment should change.
Criticism 5: retrospective diagnosis of historical figures is speculation. The book discusses several figures with apparently autistic traits (Henry Cavendish, the eighteenth-century chemist famous for extreme social avoidance and single-minded focus, among them). These passages are compelling and methodologically indefensible: you cannot assess the dead, and you cannot safely project today's criteria back into a different system of social norms. Fine as a narrative device; not usable as evidence.
And a note on how to read it: this is narrative journalism, not a review article. It moves on people, scenes and reversals, and it is extremely readable — but if what you want is the current state of the genetic evidence, or which interventions have randomised trial support, the book does not supply that and does not pretend to. It is trying to replace your frame, not your clinical handbook.
1. The "autism epidemic" is mostly not a surge in cases but a ruler that was widened several times: looser criteria, higher awareness, and services tied to a diagnosis account for the bulk of the rise.
2. The hardest evidence is diagnostic substitution — as the autism column rose, "mental retardation" and "learning disability" fell in step. The same children, filed in a different drawer.
3. The one-second analogy: when schools stopped forcing left-handed children to switch hands, measured left-handedness shot up. Nobody hunted for an environmental cause, because everyone knew the counting rule had changed.
4. The field was hostage to a narrow definition for forty years: Kanner's 1943 rare severe syndrome beat Asperger's contemporaneous continuum, at the cost of generations misfiled and missed.
5. Unknown cause + a presentation that depends on interaction = the person closest to the child with the least standing gets named as the cause. "Refrigerator mother" rested on proximity, not evidence — and it was demolished by one of the accused, Rimland, the father of an autistic son.
6. The curve really starts with Lorna Wing: Camberwell found few children matching Kanner and several times as many with the same core difficulties, so she widened the door and built the spectrum. Which is a multi-dimensional profile, not a mild-to-severe slider.
7. The vaccine panic needed three things at once: two curves coinciding in time, an onset window that lands on the vaccination schedule, and an explanatory vacuum. In a vacuum, a nameable culprit always beats "we don't know."
8. Neurodiversity's model swap is one question: is it the legs or the stairs? Open-plan noise, speech-only interviews, mandatory eye contact — all stairs, and stairs are far cheaper to replace than legs.
9. The tribe had to wait for the internet because asynchronous text happened to rebuild sociality around the way autistic people process. Whether a group forms depends less on its numbers than on whether it has a medium it can use.
10. In practice the change is to the objective function: from "make him indistinguishable" to "make him able." And in the same breath, honestly: the book's defence of Asperger was overturned by the 2018 archives, and the nonspeaking, lifelong-care end of the spectrum still struggles to find a place in the difference-not-deficit story.