The Theory
Produced the Results

The history Uta Frith left out — Sukhareva first, Asperger's documented referral history, the erasure of the women who built the foundational knowledge now being defended, the diagnostic substitution that reclaimed lost people rather than manufactured new ones — is documented in The Fairy Tale Version. This document goes somewhere the history alone cannot: into the theory, the money, the mechanism, and the law. It is built to be read by anyone who wants to challenge the argument. Every claim has a source. Every source is named.

This document was produced outside institutional research infrastructure — without academic affiliation, research team, or grant funding. The sources are peer reviewed. The argument is falsifiable. It follows a pattern the document itself traces: community knowledge preceding institutional recognition by years or decades. Sukhareva 1925. Collective neurodiversity theory developed in online autistic communities from 1992, confirmed by Botha et al. 2024. Community identification of masking, early mortality, and medication sensitivity before peer review confirmed each. The knowledge existed before the funding reached it. This document does not catalogue the total output of autism research. It examines the relationship between what was funded, what was found, and what reached autistic people's lives. That scope is intentional.

The funding record is the field's own document. What it shows is documented below.

This analysis does not argue for a single causal mechanism. It documents a consistent cross-domain pattern in which theoretical framing, funding allocation, institutional path-dependence, and intrinsic research constraints interact to produce the observed outcome profile. The question is not whether each factor exists independently — all are documented. The question is whether the pattern, taken together, is consistent with the claim that community advocacy is the primary explanatory variable. The funding record answers that question.

Frith's concern about diagnostic validity is not invented. The lumpers-and-splitters problem in nosology is real. When a diagnostic category loses predictive utility — when it no longer reliably predicts need, prognosis, or intervention — it stops functioning as a medical tool. That is a legitimate scientific question and it deserves a serious answer.

The serious answer is that the diagnostic expansion has multiple documented causes: reclassification of people previously lost under other labels, expanded search in populations the original criteria were never designed to capture, DSM-5 administrative consolidation that folded separately-counted diagnoses into one umbrella, and growing awareness in demographics the field spent decades ignoring. The data supports each of these. What it does not support is the claim that advocacy corrupted the science.

The numbers Cory Franklin cites as evidence of explosion require disaggregation before they mean anything. King and Bearman's 2009 study of California's caseload found that approximately 24% of the increase in autism diagnoses was attributable to diagnostic substitution — people previously classified under intellectual disability being reclassified as criteria evolved. Shattuck's analysis found that between 1994 and 2003, the mean increase in the autism category across US special education data was almost exactly matched by a corresponding decrease in intellectual disability and learning disability classifications. These were not new disabled people. They were the same people, finally under their correct label.

The children previously institutionalised as feeble-minded, mentally defective, or subnormal were always part of this population. The field just refused to see it until Lorna Wing named the spectrum — and even then, the resistance was institutional, documented, and sustained. Diagnostic substitution is not a concession. It is evidence that the original narrow criteria were wrong about who belonged. That is an argument against rolling back the diagnosis, not for it.

The sorting mechanism did not cease when the diagnostic categories changed. The evidence suggests it redirected. Black and Indigenous children remain significantly more likely to receive a conduct disorder diagnosis than an autism diagnosis — a label that routes toward punishment rather than support, toward the justice system rather than the clinical one. Mandell et al.'s 2007 study found that African-American children were approximately 2.6 times less likely than white children to receive an autism diagnosis on their first specialty visit, and approximately twice as likely to receive a conduct disorder diagnosis instead. The diagnostic substitution literature documents what was reclaimed when the categories corrected. The racial disparity literature documents what the correction did not reach. These are not separate problems. They are the same sorting mechanism operating on different populations through different administrative instruments.

The intersection of Indigeneity and disability adds a further dimension the diagnostic expansion debate has not adequately engaged. Indigenous frameworks for understanding difference, cognition, and community membership do not map onto Western diagnostic categories — and the imposition of those categories onto Indigenous children carries a documented colonial history that shapes both who gets assessed and what the assessment produces. Patton's 2022 analysis of disability as a colonial construct documents how Western conceptualisations of disabled Indigenous children have historically operated to pathologise cultural difference rather than identify genuine support need. Yellowhorse's 2023 work and Bevan-Brown's research on Indigenous perspectives on disability and inclusion establish that the communities most underserved by the current diagnostic system are also the communities whose knowledge frameworks have been most systematically excluded from shaping it. The argument that the spectrum has become too inclusive does not engage with the populations for whom the spectrum remains structurally inaccessible. That silence is not neutral.

The evidentiary record on diagnostic disparity reframes the expansion argument in a way the overdiagnosis framing cannot accommodate. If the diagnostic system were producing too many autism diagnoses indiscriminately, the disparity between racial groups would not be as consistent or as directional as the published literature shows. Overdiagnosis distributes errors randomly. Systematic underdiagnosis of specific populations — consistently routing them toward punitive rather than clinical frameworks — is a pattern with a direction. That direction is documented. The argument that the spectrum has become too inclusive is difficult to sustain alongside evidence that for significant portions of the population the spectrum remains structurally out of reach. Both cannot be true simultaneously without acknowledging that the system is not simply expanding — it is expanding unevenly, along lines that reflect existing hierarchies of whose presentation is recognised, whose history is taken, and whose account of their own experience is believed.

There is a version of this argument that deserves a direct answer. Fix the funding AND tighten the criteria simultaneously — the two are not mutually exclusive. Address the resource crisis without restricting diagnosis. It is a reasonable position and it has a reasonable answer. A properly resourced formal clinical tier — with mandatory intersectional screening for co-occurring conditions, auditable clinical records, and a needs-based support assessment as the output — makes diagnostic restriction unnecessary. The gate already exists. The formal clinical tier already requires clinical documentation, developmental history, differential diagnosis of co-occurring conditions, and professional assessment. The problem is not that the gate is too wide. The problem is that the gate is under-resourced, inconsistently applied, and operating on dramatically longer waiting times than a decade ago. The solution to a broken gate is to resource it properly — not to narrow it so fewer people can reach it. The people who will be screened out by a narrower gate are not the people currently waiting years for an NHS assessment. They are the people with the highest support needs and the least capacity to navigate an even more restricted system. Diagnostic restriction as the solution to a resource problem resolves the problem by reducing the entitled population. That is a political choice. It is not a clinical one. And it creates a methodological crisis that the funding argument doesn't solve — restrict the diagnosis and you produce false negatives at scale. People below a narrowed threshold do not return to anything. There is no "back to" — because there was never a complete and functional diagnostic picture waiting for them in those prior categories. They do not go back to BPD; they get a BPD diagnosis that does not explain their sensory profile, their communication differences, their executive function breakdown, their inability to sustain employment in an environment never designed for them. The correct differential was never performed — because the correct differential requires autism to be on the list. Remove it from the list and you have not helped them. You have removed the one framework that might have generated a correct and complete clinical picture. What actually happens is a referral circuit: mental health, hearing, ophthalmology, back to mental health. The intersectional picture is never assembled because no single clinician in that circuit has the mandate or the tool to assemble it. And when that circuit fails repeatedly — which it does, at scale, with documented consequences — the person is not stabilised. They are failed until they stop trying. Or they locate a community that correctly identifies what was missed, and they advocate from there with the exhaustion of someone who spent a decade being told the problem was them. That is not a fringe outcome. That is the predictable result of a restricted diagnostic system encountering the population it has decided to exclude. The restricted dataset then gets used to study the restricted population and produces findings calibrated to the restriction — because the people who would have disproved it are no longer in the data. That is not scientific refinement. That is a methodology producing its own confirmation. Narrow the gate and the foundation gets worse.

The history of how the field's foundational sample was constructed — Kanner's eleven clinic-referred cases in 1943, the exclusion of Sukhareva's prior work, the canonisation of a framework with documented connections to Nazi-era child euthanasia referrals — is documented in full in The Fairy Tale Version with primary sources. This section records the timeline for reference and builds from it.

One objection to the Sukhareva argument deserves direct address before it is deployed. Eugen Bleuler coined the term "autism" in 1911 — fourteen years before Sukhareva — from the Greek autos meaning self, to describe a symptom he observed in schizophrenic patients: withdrawal into an inner world. The word predates Sukhareva. The condition does not. Bleuler was naming a characteristic within his schizophrenia framework. He was not describing a distinct neurodevelopmental condition in children with its own clinical profile, developmental trajectory, and presentation independent of psychosis. Sukhareva was. Kanner borrowed Bleuler's term in 1943 — explicitly — because the self-absorbed quality seemed to fit what he was observing. He took the word. He did not inherit the clinical description. The claim this document makes is precise: Sukhareva gave the first clinical description of autism as a distinct neurodevelopmental condition in children — the thing the DSM-5 now describes — in 1925. Manouilenko and Bejerot confirmed in 2015 that her descriptions match DSM-5 criteria point for point. Bleuler's 1911 symptom descriptor does not touch that claim. Raising it as a counter-argument conflates the coining of a word with the identification of a condition. The document knows the difference.

Bleuler 1911 coins a word describing a schizophrenia symptom in adults. Kanner 1943 borrows the word, applies it to children, gets called the father of autism. Asperger 1944 describes similar children — to sort them for experiment or death — and gets a syndrome named after him. Frith 1991 translates Asperger, champions his framework into the anglophone canon.

Sukhareva 1925 describes the actual condition, in children, explicitly separate from schizophrenia at a time when the field had not made that distinction — the children who came before her would have been folded into broad psychopathic categories, classified as pre-schizophrenic under Bleuler's framework, institutionalised without diagnosis, or in the case of girls who masked well enough, invisible to any classification system entirely. She documented the full range, including the girls, including the subtler presentations, including the social camouflage, matching what the DSM-5 now describes point for point. That separation — autism is not schizophrenia, it is a distinct constitutional condition, stable, not a precursor to psychosis — took Western psychiatry until DSM-III in 1980 to formalise. Sukhareva made it in 1925. She was a Jewish woman working in Soviet-era psychiatry writing in Russian.

Gets erased.

The canonical history of autism has a male genealogy not because the men got there first but because the woman who did was erased from it. The diagnostic restriction argument reaches for Kanner. The plural autisms framework reaches for Asperger. The profound autism campaign reaches for both. Sukhareva is not cited in any of these arguments.

One further precision matters before the timeline. Bleuler named autism as a symptom of schizophrenia in 1911. That symptom — withdrawal into an inner world, self-referential focus — still exists within schizophrenia diagnosis today. It was not renamed or removed when the conditions were separated. It is still there. Sukhareva's contribution was not to dismiss Bleuler's symptom. It was to identify children presenting with that symptom and say — these children are not pre-schizophrenic. This is a distinct constitutional condition. The symptom overlaps. The condition does not. That clinical precision — made in 1925, formalised by Western psychiatry in 1980 — is still not fully resolved. The standard diagnostic tools including the ADOS do not consistently separate autism from schizophrenia spectrum in adults. Autistic people are 3.5 times more likely to receive a schizophrenia diagnosis than the general population. The overlap runs primarily through negative symptoms — flat affect, social withdrawal, reduced speech — the same presentation the deficit framework consistently misreads. Before systematic screening existed, autistic people were found inside schizophrenia wards, schizophrenia diagnoses, schizophrenia frameworks. Some still are.

Autism and schizophrenia are neurological siblings — not the same condition, not requiring the same clinical response, not to be collapsed diagnostically. The distinction is real and clinically important. But they share documented genetic risk pathways and overlapping phenotypic features — confirmed across multiple meta-analyses including Lugo Marín et al. 2018, De Crescenzo et al. 2019, and Lai et al. 2019. They present differently. They need different clinical responses. The distinction is real and clinically important — the wrong diagnosis produces the wrong treatment and Oliver McGowan is the consequence of that in one direction. But they share a neurological lineage the diagnostic system has artificially separated. The diagnostic system has historically attached different levels of stigma and different accommodation infrastructure to each label — not because the neurology warranted different treatment but because the institutional responses developed differently. One has a community that organised and built rights frameworks and fought for decades to have its neurology understood rather than managed. The other is still waiting for the infrastructure to catch up. The rights floor has to extend to the whole family or it isn't a floor. It's a club. And the drug sensitivity research, the immune response research, the pharmacological profile work — none of it produces valid results without full spectrum representation. You cannot study the neurological family by excluding half of it and call it rigorous.

Weak Central Coherence and Theory of Mind — the cognitive frameworks most associated with Frith's work — became the intellectual architecture that organised three decades of autism research. They were not just academic positions. They were the theoretical foundation that determined what questions got asked, which populations got studied, and where billions in research funding were directed.

Neither framework is settled science. Theory of Mind research has faced substantial replication challenges. Studies using non-verbal or adjusted methodologies have found autistic performance far less impaired than the original paradigm predicted. Autistic researchers — including Morton Ann Gernsbacher, Michelle Dawson, and Damian Milton — have systematically challenged the deficit framing, demonstrating that what the framework categorised as impairment was frequently a measurement artifact: the tool measuring the wrong thing in the wrong context.

Milton's double empathy problem reframes the entire paradigm. The difficulty is not a one-directional deficit in autistic social cognition. It is a two-directional communication difference that operates across neurotypes — non-autistic people are equally poor at reading autistic people as the reverse, but only one group gets pathologised for it. This is not a fringe position. It is published, peer-reviewed, and has been building in the literature for over fifteen years.

The 2007 Dawson et al. study is the sharpest evidentiary point. Autistic children scored on average 30 percentile points higher on non-verbal reasoning tests than on standard IQ scales. A third of children labelled low-functioning scored in the average range when tested differently. The DSM-5 itself acknowledges that IQ scores in autism are unstable, particularly in early childhood. The tool was measuring the framework's assumptions, not the child. And the framework's assumptions were built on the 1943 sample.

It matters because if the theoretical baseline was wrong, everything built on it — the research questions, the interventions, the funding priorities, the diagnostic criteria — was answering questions that were already contaminated at the foundation. The problem is not that advocacy introduced bias into a clean system. The problem is that the system was built on a biased sample, generated a biased theory, and then used that theory to determine what counted as a legitimate research question for three decades. The argument is not that theory alone determined every outcome. Theory operated within a system of institutional inertia, path-dependent funding, intrinsic measurement challenges, and competing sources of bias — all real and documented. The argument is that theoretical framing constrained the question space, shaping which signals were pursued and how heterogeneity was interpreted, in a condition where measurement tools were unreliable and the population being studied was excluded from the knowledge production process. That constraint is not a footnote. It is a mechanism — and it interacted with every other factor to produce the pattern the funding record documents.

The deficit model requires a normal to deviate from. There is no normal. There is only an average — a statistical description of a distribution, not a biological standard, not a clinical target, not a rights-bearing category. The pathologization apparatus does not exempt neurotypical people — the DSM has expanded steadily across the full range of human experience, catching grief, shyness, childhood energy, and ordinary variation in its diagnostic net. But it applies asymmetrically. The bell curve has two tails and the system has a direction. Deviation that disrupts institutional function gets called a disorder. Deviation that serves it gets called a gift, or gets called nothing at all, because nothing needs to be done about it. The child who cannot sit still gets a diagnosis. The child who learned to read at three does not — not because their neurology is less unusual but because their deviation does not require the institution to respond. The sorting mechanism is not neurotypical versus neurodivergent. It is institutional utility as the filter for what gets named, what gets managed, and what gets left alone. A model built on that filter is not measuring neurology. It is measuring friction. And it is calling the friction the condition. Mottron et al.'s Enhanced Perceptual Functioning model — published in 2006, built on the same biology the deficit framework claimed to own — shows that what WCC framed as a deficit in global processing is more accurately described as enhanced local processing. Same neurology. Different frame. One assumes the neurotypical average is the standard against which all processing is measured and deviation is pathology. The other describes a different cognitive profile with its own architecture, its own strengths, and challenges that are largely produced by a world built for a different profile. Harvey Blume wrote in 1998 — before the plural autisms framework existed as an academic proposal, before the journal submissions, before the partial founder credits — that neurodiversity may be as crucial for the human race as biodiversity is for life. Who can say what form of wiring will be best suited to the next set of challenges. The community described this from lived experience before the peer-reviewed literature confirmed it. The timeline is documented. Botha et al. 2024 establishes collective neurodiversity theory emerging from online autistic communities from 1992 — predating the academic frameworks that now claim partial co-authorship.

Neurodiversity already solved the problem the plural autisms framework is now trying to solve with new diagnostic categories. Human neurology is plural. Different profiles, different presentations, different support needs — this is not a discovery requiring new taxonomic apparatus. It is what the community has been saying since the community had a means to say it. The field's response has been to taxonomise the variation rather than accommodate it. The plural autisms framework applies a taxonomic apparatus to this reality only when it occurs in people the system was not built for. Co-occurring conditions are not a feature unique to autism. They are a feature of being human. The sorting mechanism appears only where the system needs to manage the people it finds difficult to accommodate.

The heterogeneity the plural autisms framework identifies is real. It is documented at every level of analysis and the document does not dispute it. Genetically, a 2025 analysis identified four biologically distinct programs with different developmental trajectories and clinical profiles — genuine variation in the underlying architecture, not measurement noise. The finding is unreplicated. What it confirms — that autism is plural, that presentations differ meaningfully, that a single category encompasses neurologically distinct profiles — the autistic community documented before the genome caught up. The confirmation is welcome. The framing as breakthrough reproduces the attribution problem this document has already named: the community said we are plural decades before the research caught up, and the research that confirms it gets the credit. Four biological subtypes do not produce four different diagnoses. A person with subtype 2 biology presenting autistically is autistic. The gene adds precision to the mechanism. It does not fracture the category. Co-occurring conditions intersect with autistic neurology in ways that produce genuinely different presentations and support needs — intellectual disability, FASD, epilepsy, catatonia, autoimmune conditions each changing what the neurology looks like from the outside without changing what it is from the inside. And intersecting identity — race, gender, class, geography, cultural context — shapes not just who gets diagnosed but how the neurology expresses itself in a world built to accommodate some people and not others.

Research has identified consistent patterns in autistic brain development across multiple studies — including differences in synaptic pruning rates, local overconnectivity within brain regions, and underconnectivity between distant regions — that provide neurological grounding for the cognitive and sensory differences the autistic community has documented. These patterns are not universal across all autistic people and the research is ongoing. No absolute conclusions follow from them. What the consistent findings do establish is that the differences are neurological in origin and that the interaction between those differences and environmental design produces the disability. That is the social model stated in neuroscience. Autism, like all diagnostic categories, is also a social construct — the category, its boundaries, and the meaning made of neurological difference are shaped by the frameworks, values, and power structures of the institutions that define them. The neurology is real. What the field does with it is a values question, not a scientific one. Science has a responsibility not to encode one world view as neutral fact while presenting itself as objective. The documented pattern — whose neurology gets studied, whose experience gets centred, whose differences get pathologised and whose get accommodated — is not a scientific finding. It is a political choice that the scientific framing has made harder to contest.

Plural is correct. The framework built on it is not. The plural autisms proposal resolves heterogeneity by drawing more lines inside the category. Neurodiversity resolves it by correctly locating the category within a broader understanding of human neurological variation. The cultural dimension is the one the plural autisms framework is least equipped to handle. Autistic presentation in a high-masking-demand cultural context looks different from autistic presentation in a low-masking-demand one — not because the neurology differs but because the environment is producing different behavioural outputs from the same neurological input. A diagnostic framework that reads behavioural output as the primary indicator of the underlying condition will produce different diagnoses from the same neurology depending on the cultural context it encounters. That is not evidence of plural autisms. That is one neurology in plural environments. The research methodology that cannot distinguish between those two things does not have the precision required to be splitting diagnostic categories on the basis of what it finds. It is mistaking the environment's effect on the presentation for a feature of the neurology itself. That is the 1943 sample problem restated at a global scale.

Neurodiversity was always the correct umbrella — not because it flattens the variation but because it names it correctly. Human neurological variation is real, plural, and the variation itself is not the pathology. Autism is one cluster within that variation — internally plural, shaped by genetic architecture, co-occurring conditions, cultural context, and intersecting identity — sitting within a broader neurological family that shares genetic risk pathways and overlapping phenotypic features across conditions the diagnostic system has sorted into separate administrative categories. ADHD, dyslexia, dyspraxia, the schizophrenia spectrum — these are not separate populations accidentally sharing features. They are a neurological family the diagnostic system encountered at different moments in different institutional contexts and named separately without understanding the shared architecture underneath. The community described this from the inside before the genome confirmed it. The plural autisms framework proposes to resolve this by adding more categories. The neurodiversity framework proposes to resolve it by understanding what the categories are describing. One produces more paperwork. The other produces better science.

Claiming co-authorship of a framework the community developed and described for decades before the first journal submission is not a scientific contribution. It is attribution without precedence. Neurodiversity offered a framework that could have redirected thirty years of research toward the questions that actually matter to autistic people's lives. The funding record shows which framework got the money. The outcome data shows what that choice cost.

The foundation was wrong. Not wrong in every finding — wrong in what it could see. Of course what it couldn't see didn't reach the people who needed it.

Kanner's 1943 sample produced a biased theory. The biased theory produced a biased research agenda. The biased research agenda produced biased tools — diagnostic instruments, IQ measurements, assessment protocols — that couldn't reach the people they were supposed to assess. The medication protocols didn't account for atypical responses. The clinical training didn't account for what autistic neurology looks like under pharmacological load. Oliver McGowan died in 2016 because clinicians administered antipsychotic medication against his and his family's explicit wishes, despite his documented history of adverse reactions. An independent review found his death potentially avoidable. His mother Paula McGowan campaigned until mandatory learning disability and autism training became law in the UK Health and Care Act 2022 — now required for every health and social care worker registered with the CQC. In 2022, 42% of deaths of people with a learning disability were rated as avoidable compared to 22% for the general population. Oliver McGowan's case is not an outlier. It is the foundation error expressed in a clinical encounter. The wrong theory produced the wrong tools. The wrong tools produced the wrong training. The wrong training produced avoidable deaths. And the proposed solution — sorting the population further using the same tools built on the same wrong foundation — does not address the foundation. It extends it.

The dominant deficit-based framing has shaped what gets measured, how results are interpreted, and which signals are considered meaningful. In a heterogeneous condition like autism, that increases the likelihood of weak, non-replicable findings and limits clinical translation. The biomarker replication crisis reflects multiple factors — heterogeneity, small samples, measurement noise — but framing is a major and under-acknowledged contributor. Subtyping within the same deficit framework does not resolve this. Sorting a spectrum into smaller categories using the same unreliable tools asks the same wrong question about a smaller population. The spectrum remains heterogeneous regardless of where the diagnostic boundary is drawn. A complementary research program is needed: not only asking what caused these profiles, but systematically investigating what these neurologies do, how they process information, and under what conditions they function well or break down. The field funded the first. The community needed the second.

The Interagency Autism Coordinating Committee — the federally mandated body that includes autistic community members, researchers, and government agencies — formally recommended in 2017 that funders prioritise treatments and interventions, evidence-based services, and lifespan issues. These were not preferences listed alongside other options. They were the stated top three priorities, formally recorded in the Strategic Plan, the document the IACC exists to produce. The process that produced them was itself incomplete — autistic self-advocates gained partial voting representation late, capped at one third of public members under the Autism CARES Act reauthorisation, a proportion the funding record confirms did not produce CRPD-compliant community direction, and the broader community that had emerged through internet connectivity represented a population the formal process had not systematically captured. That broader community — late-diagnosed adults, women who recognised themselves outside clinical settings, people from demographics the original criteria excluded — consistently identified the same priorities the formal process recorded. The convergence matters. The betrayal runs through both. Even the priorities recorded through that imperfect process were only partially funded — interventions received some allocation while services and lifespan research, the two priorities the community had most consistently identified as critical, received the smallest fractions of the entire portfolio.

Here is what the money did instead.

The IACC's own 2019–2020 Portfolio Analysis confirms the pattern did not change. Biology comprised 45.3% of total funding in 2019 ($192.3 million) and 45.4% in 2020 ($190.3 million). Of the $418.9 million total autism research spend in 2020, 8.4% went to services research and 4.3% addressed the needs of autistic adults. For research specifically focused on autism in girls and women in 2020, biology accounted for 88.3% of funding. Services, interventions, and screening for girls and women had zero dedicated projects that year.

The Ohio State University research team that examined these figures asked the question that nobody in the funding apparatus has answered: "If the IACC's advice on how resources should be allocated is ultimately ignored by federal funding agencies, what is the point of continuing to have this committee make budget recommendations?"

That is not a community advocate asking. That is a peer-reviewed study published in the journal Autism in 2021. Their conclusion was direct: the IACC's recommendations had not made a substantive difference to the types of grants that actually received federal funding. The community was asked. Their priorities were formally recorded. The money went elsewhere. For over a decade. This is documented in the government's own published reports. It is not contested. It is the official record.

Science requires a purpose beyond its own perpetuation. The stated purpose of autism research — in every funding application, every grant justification, every congressional hearing that authorised the billions — was to improve autistic lives. The outcome data is the measure of whether that purpose was kept. It is above. The implicit promise of the biology funding was that understanding the cause would eventually produce actionable improvements. After decades of disproportionate investment that promise has not been kept. Biological research has produced heritability estimates and rare variant findings — copy number variants, polygenic risk scores, pathway insights. None have reached a clinical tool, a diagnostic instrument, or an intervention that measurably changed an autistic person's life at scale. The investment was real. The translation wasn't. No single genetic marker identified. No clinically actionable causation finding. The condition as genetically complex as it was at the start. Meanwhile the life expectancy gap went unconfirmed for decades. The BPD misdiagnosis chain ran for a century. Oliver McGowan died because the clinical training didn't exist. Scurvy returned in wealthy countries. Families skipped meals. The funding pattern shows a sustained emphasis on internal research continuity. The gap between investment and population-level outcome is documented in the field's own published record.

The research funded most heavily produced the fewest actionable results. The research funded least — or not at all until community pressure forced it onto the agenda — addressed the questions that most directly affected autistic people's lives and deaths. This pattern is consistent with institutional path-dependence, where existing research programs continue to attract funding independent of outcome alignment. The field disproportionately funded one line of inquiry while underfunding complementary questions that more directly relate to lived outcomes.

The most concrete demonstration of what the misallocation costs arrived in a 2026 paper in BioPsychoSocial Medicine. A retrospective study of 13 children presenting with scurvy — a disease considered eradicated in wealthy countries — found that 92% had confirmed or suspected autism. All had eating disorders. 85% had multiple nutrient deficiencies. 62% were misdiagnosed by their previous doctor. The solution was a tasteless, odourless micronutrient powder — an environmental accommodation. While this is a small retrospective sample, it illustrates how environmental mismatch can produce severe outcomes independent of underlying biology. These children developed a medieval disease not because their neurology is deficient but because the food environment was built for one sensory profile and refused to accommodate theirs. The research portfolio that spent 45% of its budget on biology and 2.5% on quality of life and services did not prevent a single case of scurvy. An accessible food environment would have. The question of which research gets funded is not a scientific question. It is a values question. And the values question has an answer in the outcome data.

The outcome data at the family level is documented by the Joseph Rowntree Foundation and Family Fund's Cost of Caring 2025 report. Disabled people in the UK have a poverty rate of 28% compared to 20% for non-disabled people. Among families raising disabled or seriously ill children: 44% say their income does not cover basic day-to-day essentials. 87% have no savings at all. 51% of parents are skipping meals or cutting portion sizes. A third cannot afford to keep their home warm. Very deep poverty is at record levels, affecting 6.8 million people — nearly half of everyone living in poverty. Destitution more than doubled between 2017 and 2022. This is the downstream product of the funding record. Decades of biology research that produced no clinically actionable results. Decades of services and lifespan research at 2.5% of the portfolio. The families in these figures are not a policy abstraction. They are the people the research was supposed to serve. The gap between what was funded and what was needed is measured in skipped meals and cold homes and medieval diseases returning in wealthy countries.

The resistance to community-directed research was not passive — it was documented in peer review. AASPIRE, the Academic Autism Spectrum Partnership in Research and Education, has been conducting participatory autism research since 2006 with autistic adults as co-investigators from design through dissemination. When they submitted early proposals, peer reviewers questioned whether autistic adults could serve as research partners at all. When AASPIRE demonstrated successful inclusion, reviewers responded by assuming the autistic partners must not be "truly disabled." This is the field's documented reaction to the methodology that best serves the population it studies — recorded in the peer review process, the gatekeeping mechanism of the research establishment itself. The community was not corrupting the science. The science was actively resisting the community's participation in it.

The Interagency Autism Coordinating Committee was established under the Combating Autism Act 2006. Its mandate is to coordinate federal autism activities and make recommendations to the Secretary of Health and Human Services on research priorities. It includes federal agency representatives, researchers, autistic self-advocates, and family members. Its annual community input process formally records what the community identifies as priorities. The gap between those priorities and actual funding allocation — documented in Section 04 — is the IACC's own evidence that its process does not work.

The fight to include autistic self-advocates as voting rather than observing members on the IACC is a documented record of institutional resistance to community direction. The Autism CARES Act 2014 strengthened requirements for autistic representation. As of the most recent reauthorisation, still only one-third of the 21 public members are required to be autistic — a composition ASAN explicitly identified as insufficient. The committee that exists to set priorities on behalf of the community does not require the community to hold the majority of its seats.

The billions directed toward biological research created careers, institutions, grant pipelines, and journal prestige all calibrated to a specific theoretical framework and population sample that was never representative. The people most invested in the narrow model are not just ideologically attached to it. They are financially and institutionally attached to it. A reorientation toward community-directed, outcomes-focused research requires not only new priorities but a reassessment of whether the questions that consumed the previous allocation were the right ones given the outcome profile that resulted.

The "collapse" now being identified is consistent with the predictable consequences of a framework encountering the population it was not designed to include — a population the original diagnostic criteria, shaped significantly by the theoretical frameworks this document examines, were not built to capture.

The move is not finished. While one position contracts the spectrum from the top — too broad, too inclusive, lost its meaning — a parallel campaign is working from the bottom. The "profound autism" push, which nearly secured dedicated subcategory language in the most recent Autism CARES reauthorisation before being pulled back after community pushback, seeks to formally split the single spectrum category that Lorna Wing built specifically to prevent sorting by support need.

The proposal fails on its own clinical terms before it reaches the funding argument. Under DSM-5, ASD remains a behavioural diagnosis requiring specific social-communication deficits and restricted, repetitive behaviours to be present and reliably assessed above any global impairment. In individuals with profound developmental disability, that assessment is frequently impossible. If the behaviours cannot be reliably identified, ASD cannot be diagnosed. As a result, many of the most profoundly disabled individuals — the population the campaign claims to centre — are already excluded from autism research portfolios not because the spectrum is too inclusive but because the assessment tools cannot reach them. A "profound autism" subcategory inherits exactly this problem. The diagnostic barrier does not move. The population the campaign describes remains outside the clinical reach of the new label for the same reason they are outside the current one.

The funding data makes this worse. Although DSM-5 consolidated all former Pervasive Developmental Disorder categories into ASD, this diagnostic merger did not produce equitable research investment. Autism receives approximately $305 million annually from NIH — one of the largest coordinated childhood neurodevelopmental research portfolios in existence. The high support needs population was already inside that category. They received almost none of it. Forty-five percent went to biology. Two point five percent went to lifespan. The conditions most commonly co-occurring with high support needs autism — global developmental delay, intellectual disability, FASD, epilepsy, complex metabolic disorders — have no dedicated NIH funding lines at all. Global developmental delay has zero. No dedicated category. No coordinated research mechanism. No trackable investment. The profound autism campaign is proposing to solve a funding allocation failure with a diagnostic category. That is the wrong tool for the problem. Ballreich et al.'s 2021 analysis of NIH funding across 46 disease categories confirmed that NIH funding is path-dependent and does not follow disease burden. Creating a new subcategory does not create new money. It moves the label while the funding void the label was supposed to address remains exactly where it was.

The allocation failure has a documented lineage. The $305 million portfolio was built on the theoretical framework that followed Asperger's presentation — verbal, high-functioning, male. The spectrum became broad enough to include everyone. The funding remained narrow enough to serve the Asperger end of it. The high support needs population was nominally included and practically ignored. The profound autism campaign now uses that practical ignorance as the argument for a split that would make it structural and permanent — while removing the political coalition that gives everyone access to shared rights frameworks, legal protections, and a research infrastructure that could, if redirected, actually serve them. The solution is not a new label. The solution is a redirected allocation and an intersectional assessment infrastructure that addresses the co-occurring conditions driving support complexity. Australia demonstrated that structural community direction changes the allocation within five years. The profound autism campaign is proposing to solve in a decade with a new diagnostic category what a funding redirect could accomplish in five years with existing infrastructure. Diagnostic consolidation did not produce research equity. It produced a larger label applied to the same narrow allocation. The answer is equity in the allocation. Not a return to sorting.

The profound autism campaign rests on a linear model of the spectrum that the neuroscience does not support. The neuroimaging literature confirms that the autism spectrum is not a straight line from low support needs to high support needs. Brain heterogeneity across the spectrum is documented as non-linear and dynamic across the lifespan. Support needs fluctuate — they increase under stress, illness, sensory overload, accommodation withdrawal, and life transition. They decrease with appropriate support, environmental adjustment, and reduced demand load. An autistic person whose support needs spike dramatically during a medical crisis, a bereavement, or a systematic removal of accommodations is not regressing into a different category. They are an autistic person whose nervous system is responding to load in the way autistic nervous systems respond to load. The profound autism campaign treats that fluctuation as a stable categorical distinction. The longitudinal data does not support that treatment. Support needs as a framework — person-centred, dynamic, responsive to environment — is both more clinically accurate and more rights-consistent than a diagnostic split that encodes a snapshot as a permanent category. The question the campaign never answers is: what happens to the person whose support needs are profound at 35 and moderate at 45 because their environment finally changed? The split has no mechanism for that person. The spectrum does.

The profound autism campaign does not replicate Asperger's framework directly. It does something structurally more significant. It takes the inclusive framework Sukhareva documented and Lorna Wing formalised — the full range, everyone belongs, one rights floor — and proposes to reintroduce the support-need sorting Wing explicitly built the spectrum to prevent. Wing drew the spectrum wide because she understood what diagnostic sorting by support level produces historically and clinically. The campaign proposes to undo that from both ends simultaneously. From the top, late-diagnosed women, online community members, and self-identified autistic people are characterised as probably anxious or diagnostic boundary cases. From the bottom, people with the highest support needs — the population Wing specifically brought into the spectrum — are proposed for a separate subcategory. What this squeeze preserves in the middle is a population closer to Asperger's original clinical selection: verbal, assessable by existing tools, accommodatable within current infrastructure. The document does not characterise intent. It characterises structural outcome. The proposal reintroduces the sorting Wing rejected, using the framework she built to reject it. The people most affected by this structural outcome are the least positioned to participate in the policy debate that determines it. That asymmetry is not incidental to the proposal. It is a predictable feature of it.

Severity is not a neurological fact. It is an environmental measurement. What the system calls profound autism is frequently profound abandonment. Remove the AAC device. Strip the internet access. Disrupt the schedule. Withdraw the two-to-one support. Then measure what remains and call it a diagnostic category. This reflects a methodological limitation rather than a valid basis for classification — the measurement is capturing the interaction between the neurology and the stripped environment, not the neurology itself. A person classified as level one in a hostile environment with stripped accommodations can be more functionally disabled than a person classified as level three with appropriate support in place. The spikey profile means strengths mask needs. The access tool makes competence visible — remove it and the competence becomes invisible to the system while remaining entirely present in the person. Support need is a snapshot of a person in a specific environment at a specific moment. It is not a permanent neurological category. Encode it as one and you have built the measurement error into the diagnostic infrastructure. The person whose support needs are profound at thirty-five and moderate at forty-five because their environment finally changed has no mechanism in the split. The spectrum does.

The IACC betrayal is documented most comprehensively in the United States because the US has the most transparent public funding record. But portfolio analyses now exist for the UK, Australia, New Zealand, and Canada — and the pattern is the same across all of them. This is not a US policy failure. It is a global research culture failure. And it has been documented internationally with the same methodology, producing the same finding, for over a decade.

In the UK, Pellicano, Dinsmore, and Charman's 2013–2014 analysis found that 56% of all UK autism research grant expenditure between 2007 and 2011 went to biology, brain, and cognition. Services received 5%. Societal issues received 1%. When those findings were presented to over 1,600 community members — autistic people, families, practitioners — two thirds were dissatisfied. The gap was not a surprise to the community. It was a confirmation of what they already knew. And the UK's per-person autism research investment at the time was less than £6 per year.

The Autistica/James Lind Alliance Priority Setting Partnership — the gold standard methodology for community-directed research priority setting, used across conditions from asthma to dementia — asked over 1,000 autistic people, families, and clinicians for their top research questions in 2016. The resulting top 10 were entirely about mental health interventions, services, social care for adults, anxiety, education outcomes, communication support, and sensory processing. Not one of the ten priorities identified by over 1,000 community members was biological causation research. The gap between that list and what 56% of the funding was buying is the argument made concrete.

Australia is the counter-example that proves the research allocation argument — and the jurisdiction that simultaneously documents what happens when the governance principle is won on paper but lost in implementation. These are not contradictory findings. They are the same argument running on two different tracks.

When the Cooperative Research Centre for Living with Autism was established in 2013 with structural community partnership and a whole-of-life research mandate, the funding distribution shifted measurably within five years. Biology dropped from roughly half the portfolio to 27%. Services and lifespan research increased. The pattern that had held across every other jurisdiction broke — because the structural conditions changed. Community direction built into the funding architecture produced a different allocation. That is a documented proof of concept for the research argument this document makes.

The National Disability Insurance Scheme tells a different story from the same jurisdiction — and a more instructive one. The NDIS changed service delivery structurally. Individualised funding, portability, the principle of choice and control: these were real, and the Australian disability community fought for them and won them. The Independent Review of the NDIS, reporting in December 2023 after consulting over 10,000 people, confirmed what participants had been documenting for years. Community supports for all people with disability — the broader ecosystem the scheme was always designed to sit within — were never built. The NDIS became, in the Review's own phrase, an oasis in the desert: the dominant and in many cases only source of support for people who qualified, while those outside its eligibility gates received nothing, because existing community services were defunded when they weren't folded into the scheme. The scheme designed to be one part of a system became the whole system by default. The cost and eligibility pressure that followed was the predictable consequence.

The Review found that the rhetoric of choice and control was not supported by the experience of participants. Navigating the system left families exhausted and stressed. People were required to present the worst version of themselves to qualify for support. Those with the highest and most complex support needs — the population the scheme was most urgently needed for — faced the most burdensome navigation requirements to access it. Co-design with people with disability was inconsistent, and failed reforms damaged trust between the community and the administration. The lived experience community reported that service providers were favoured in the Review's own recommendations over the voices of the people the scheme was supposed to serve.

Australia therefore does not prove that the alternative works in every domain. It proves something more precise: structural community direction changes outcomes, and the absence of structural community governance allows apparatus to capture implementation regardless of what the mission statement says. The CRC changed research allocation because the governance was built into the architecture. The NDIS won the legislative rights argument and then watched the administrative layer grow until it consumed the resource it was supposed to deliver — because the governance that would have prevented that was consultative rather than structural. Same jurisdiction. Same period. Same principle. Different outcomes based entirely on whether that principle was built in or written in. The rest of the world has the research evidence. The NDIS is the service delivery warning. The lesson from both is identical: write it into the architecture. Monitor it against outcomes. The apparatus will always grow toward its own logic if the people it serves do not hold structural power over it.

New Zealand's 2023 portfolio analysis found 67% of autism research funding allocated to biology — the highest proportion of any jurisdiction studied. Community members were dissatisfied and expressed a lack of alignment with their priorities. The paper's conclusion was direct: autistic people need to be included in autism research and related funding decisions. This finding was published in 2023. The gap it documents has existed for the entire period the field has been tracking it. The documentation is not new. The gap is not new. The choice to maintain it is not new.

NIH autism research funding reached approximately $305 million in 2024 — one of the largest coordinated childhood neurodevelopmental research portfolios in the United States, with consistent year-over-year investment since the Combating Autism Act established a national framework in 2006. The Autism CARES Act has committed approximately $5.2 billion to NIH, CDC, and HRSA over two decades. By any measure of total dollars, autism is not an underfunded condition.

Autism receives 10–100 times more funding than most individual rare or terminal childhood diseases — sustained over three decades. The resource crisis Frith and Franklin describe cannot be explained by insufficient total investment. It has to be explained by where the money went within the autism budget. And the IACC's own published data answers that question: 45% to biology, 2.5% to lifespan, 8.4% to services. The crisis is not a funding shortage. It is a documented misallocation.

Ballreich et al.'s 2021 analysis of NIH funding across 46 disease categories confirmed what the IACC data already showed: NIH funding is path-dependent and does not track disease burden. The correlation between NIH funding and disability-adjusted life years was weak. The correlation between changes in funding and changes in burden was also weak. Diseases with established categories and strong advocacy maintain their portfolios regardless of relative need. Diseases without dedicated categories — including global developmental delay and profound developmental disability — remain structurally invisible regardless of their burden. Creating a "profound autism" subcategory does not overcome this structural mechanism. It requires its own advocacy infrastructure, its own research base, its own funding pipeline to be built from scratch — while removing the population from the existing $305 million portfolio that, redirected, could serve them now.

The proposed subcategory fails on one further clinical ground that the field's own literature confirms. IQ measurement is not reliable for autistic people across the spectrum. The DSM-5 states directly that IQ scores in autism may be unstable, particularly in early childhood. The Dawson et al. 2007 study — already in this document — showed standard IQ tools underestimate autistic cognitive capacity by an average of 30 percentile points. Research on the WISC-V found statistical bias specifically affecting working memory and processing speed scores in autistic children, meaning the full-scale IQ routinely underrepresents actual cognitive functioning. The standard diagnostic instrument — ADOS-2 — has no appropriate module for adolescents and adults with intellectual disability. And research consistently shows adaptive behaviour measures are more clinically useful than IQ scores for determining what high support needs autistic people actually require. A subcategory defined by an IQ floor that cannot be reliably measured, assessed by a tool with no module for the population it is assessing, sorted by a cognitive threshold the field's own manual acknowledges is unstable — is not precision medicine. It is administrative sorting built on an unreliable measurement. The correct response is to build better assessment tools, fund co-occurring condition research, and use adaptive behaviour — what a person actually needs to live — as the basis for support planning. The DSM-5 criteria, when mapped against the historical record, reflect Sukhareva's descriptions most closely. The Jewish psychiatrist from Kyiv documented the full range in 1925 and 1927 — including the girls, the non-verbal, the full spectrum of presentation. Her work was erased. His framework got the funding. The profound autism campaign proposes to sort the population she documented using tools built for the population he described.

When diagnostic classification determines access to legal rights, supports, and accommodations, control over who receives a diagnosis is control over who receives rights. This is not a philosophical concern. It is a legal one with existing instruments, and those instruments have specific things to say about both research conducted on disabled communities and the obligation to provide accommodations.

Australia's National Disability Insurance Scheme represents the most ambitious disability support infrastructure built in any jurisdiction — and a demonstration that infrastructure without genuine community direction reproduces the same failures as research funding without genuine community direction. The NDIS was designed to give disabled people choice and control over their supports. In practice, administrative overhead consumes resources that should reach participants. The assessment process is adversarial, built around what the system needs to document rather than what the person needs to live. Participants with the highest support needs face the most complex navigation burden and receive the least proportional benefit relative to their need. The pattern is identical to the IACC betrayal documented in Section 04: the community's priorities are formally recorded, the infrastructure is nominally built to serve them, and the resources are absorbed by the apparatus rather than reaching the people. Legal frameworks and dedicated funding are necessary. They are not sufficient. What determines whether they serve the population is whether the people most affected have genuine directive power over how the system operates — not consultative input that gets recorded and ignored, but structural authority that cannot be overridden by administrative convenience.

The deeper structural problem runs beneath the service system itself. The administrative apparatus of disability support — assessments, plans, reviews, appeals, coordinators, compliance frameworks — exists partly because the political consensus holds that disabled people cannot be trusted with money directly. The result is that a significant proportion of disability funding is consumed by the system built to deliver it. Research on individual budgets, self-directed support, and direct payment consistently shows that when disabled people control their own funding they achieve better outcomes at lower cost than equivalent money managed by service systems. The CRPD's supported decision-making framework — already the binding international standard — is built on the premise that disabled people are the experts on their own lives. Every service model that substitutes professional judgment for personal choice operates in tension with that standard regardless of professional intent. Most disabilities cost nothing to accommodate. Flexible scheduling. Plain language. Quiet assessment conditions. Presuming competence. Believing the person. These are attitudinal and environmental changes that carry no significant financial cost and are withheld not because they are expensive but because the system was not designed around the people who need them. The documented pattern shows administrative costs absorbing an increasing proportion of the resource relative to direct participant benefit. The field spent decades studying autistic people rather than serving them. The service systems spend the equivalent proportion administering them rather than supporting them.

The concerns about diagnostic expansion that this document addresses were not published in a peer-reviewed journal for evaluation by the field. They were published in national media — audiences of teachers, GPs, parents, and tribunal assessors. That is a policy intervention, not a scientific contribution. The transmission mechanism was chosen deliberately, and it is the one that determines what comes next for real people in real institutional settings.

When a credentialed authority tells those audiences that the spectrum has lost all meaning, that masking has no scientific basis, that sensory accommodations lack grounding, and that young women seeking diagnosis are probably just anxious — that does not produce nuanced clinical reflection. It produces permission. Permission to disbelieve. Permission to withdraw. Permission to refuse. The clinical encounter does not become a debate about diagnostic subtypes. It becomes a door that closes.

Abrupt withdrawal of accommodations is a documented crisis trigger for autistic people. This is not a contested claim. Laxman et al.'s 2019 longitudinal study — tracking 204 autistic individuals across 14 years — found that loss of services begins well before formal transition points, and that the rate of improvement in symptoms and functioning slows or stops when services are removed. Taylor and Seltzer's prior research in the same cohort showed the pattern holds specifically at the post-high school service cliff. Pulling the support that keeps someone functional does not produce an academic debate about diagnostic categories. It produces regression, shutdown, crisis presentation, and in the worst cases, irreversible deterioration in functioning. The medical record documents what abrupt destabilisation does. Public claims by credentialed authorities that young women seeking assessment are probably just anxious provide the justification for doing it to more people. The transmission mechanism runs from credentialed media intervention to clinical encounter to documented harm.

The political climate matters here. In a moment when assisted dying access is expanding in multiple jurisdictions, when disability rights protections are under active legislative pressure, when the populations most likely to be diagnosed late — autistic women, BIPOC autistic people, autistic people with high support needs — are also the populations with the least institutional protection, publicly challenging the diagnoses of those people in policy-facing media is not a neutral scientific act. It is an intervention with a predictable direction of harm.

When credentialed authority is used in policy-facing media to make claims that foreseeably destabilise a high-risk population — through channels that bypass peer review — the accountability question does not disappear because the mechanism is publication rather than clinical practice. The harm pathway runs regardless of whether a clinical licence is held. The profession built fitness-to-practise mechanisms precisely for situations where professional authority produces foreseeable harm. The equivalent accountability mechanism for credentialed public commentary is the peer review process that was bypassed. The question of accountability does not resolve because the mechanism is emeritus rather than registered.

The live transmission mechanism is documentable in real time. When a credentialed authority tells clinical audiences that young women seeking autism diagnosis are probably just anxious — the clinical encounter produces a BPD diagnosis instead. The literature on BPD misdiagnosis in autistic women is now extensive. Autistic women are diagnosed later and with greater difficulty than autistic men. The journey to diagnosis frequently includes prior mental health misdiagnoses — and misdiagnosis with borderline personality disorder is documented as particularly common. The phenomenological literature shows participants in every case recalled autistic differences since childhood that went unnoticed. The BPD diagnosis carried stigma, led to treatments that actively encouraged masking — a documented suicide risk factor in autistic people — and left participants feeling powerless to challenge a label they knew was wrong. Receiving the correct autism diagnosis was described as life-changing. The BPD diagnosis was described as overtly harmful. This is not a diagnostic error that happens randomly. It happens to the population characterised in the media coverage this document addresses as probably just anxious. It happens because the criteria were built on a different presentation. It happens because the system was not designed to see them.

There is a clinical argument that closes this misdiagnosis chain completely. BPD is not diagnosed in childhood. The DSM explicitly discourages diagnosing personality disorders before 18 because personality is still developing. The presentation requires a stable pattern across adulthood. It is by definition not a neurodevelopmental condition — it has no childhood onset, no early developmental markers, no presentation in the developmental history that precedes the adult diagnosis. Autism is neurodevelopmental. It is in the developmental record. The baby who didn't babble on schedule. The toddler who lined things up. The child who couldn't navigate the playground. The girl who watched other girls and learned to copy them so precisely nobody noticed until the load became unsurvivable at 35. All of that history is in the file if someone looks for it. BPD cannot explain a childhood. Autism can. The misdiagnosis chain runs because clinicians took the presentation in front of them and reached for the nearest available framework — without taking the developmental history that would have made the autism diagnosis obvious. They saw the adult. They missed the child. Sukhareva saw them in 1927. The field spent nearly a century diagnosing them with everything except what she already knew they had. BPD is the current iteration of that century-long misdiagnosis chain. The papers now being written suggesting female autism presentation is actually BPD are not a scientific correction. They are the same move, in updated clinical language, producing the same harm to the same population. Check the childhood.

Marsha Linehan — the creator of Dialectical Behaviour Therapy, the primary evidence-based treatment for BPD — said: "If you meet the criteria for borderline personality disorder do not tell anybody. They'll treat you differently if you do. And many mental health professionals will refuse to see you." The architect of the intervention built to treat the diagnosis was advising people to hide the diagnosis from the system built to treat them. That is not a fringe critique of BPD. That is the field's own authority on its own diagnostic category describing a system that harms the people it is supposed to serve. The BPD diagnosis, as currently applied, functions as Linehan described — not as a clinical tool that opens doors to care but as a label that closes them. The autistic women who spent years or decades with that label before receiving a correct diagnosis described it exactly the same way. The diagnostic pattern persisted despite the architect of the primary intervention describing it as actively harmful to patients.

The harm pathway from diagnostic restriction is not only misdiagnosis. It is also what happens when correct diagnosis exists but support is withdrawn. Presentation stabilises with support. Funding bodies read stability as evidence the support is no longer needed. Support is removed. The nervous system loses the scaffold. What follows is not a behavioural problem. It is a medical emergency. Catatonia — a complex neuropsychiatric syndrome characterised by motor slowing, mutism, rigidity, and withdrawal — affects between 10% and 20% of autistic people according to the clinical literature, with some studies finding rates as high as 35-37% in adolescent populations. It is frequently missed because its symptoms overlap with autism presentation and clinicians reach for the nearest available explanation — behavioural escalation, psychiatric deterioration, non-compliance. Research confirms that catatonia is a common cause of late regression in autistic adolescents, occurring in 85% of those experiencing significant late regression in one study, with stressful life events — including support withdrawal and transition — as documented triggers. The same psychiatric medications clinicians reach for when they don't recognise catatonia can trigger or worsen it — as Oliver McGowan's case documented fatally. The child who was stable is now in crisis. The crisis gets documented as behaviour. The behaviour gets used to justify restriction not restoration. The data from before the support was removed gets cited as evidence the restriction was appropriate. Nobody asks what happened to the support. That is the harm pathway made concrete. That is what the restriction produces when it reaches real children in real funding systems.

For those who hold that autism is categorically a disorder — a position this document does not require you to abandon to follow the argument — the clinical obligations that follow are worth examining carefully. A doctor's job is to help. A clinician's job is to improve outcomes. If autism is a disorder, the clinical question is: what produces poor outcomes in this population, and what produces good ones? That question has an answer. It is in the data. And the data does not point where the disorder model assumed it would.

Autistic people present on average with significantly elevated rates of anxiety, depression, autoimmune conditions, gastrointestinal disorders, sleep disorders, chronic pain, eating disorders, and suicidality. Early mortality is documented across multiple cause-of-death categories. These are not incidental findings. They are the outcome profile of a population. And they require explanation — not description, explanation. A disorder model that can describe the deficit without explaining the mechanism or the variance is not a clinical tool. It is a label. Labels do not improve outcomes. Mechanisms do.

The mechanism the population-level data points toward is not located inside the autistic person's neurology in isolation. It is located in the interaction between that neurology and the environments, systems, and clinical encounters it has been subjected to. Chronic allostatic load from sustained mismatch between neurological architecture and built environment. Immune dysregulation from chronic stress. Metabolic and neurological consequences of masking — the chronic suppression of natural responses to produce a performance of neurotypicality that the system finds less inconvenient. Iatrogenic harm from medications prescribed without understanding how autistic neurology processes them. Co-occurring conditions that were never identified because the assessment only had one question on the list. Cascading crises from support withdrawal that the clinical record documented as deterioration rather than as the predictable consequence of removing the scaffold that was doing the work.

If autism is a disorder, these are the disorder's mechanisms. They are environmental. They are systemic. They are in significant part produced by the clinical system's own failures to identify, accommodate, and correctly treat the population it was supposed to be helping. The doctor's job is to help. Helping requires understanding what is actually producing the harm. The harm is not located where thirty years of biology funding was looking. It is located in the gap between what this neurology needs and what every system the person moves through was built to provide. Close the gap. That is the clinical intervention. That is what helping looks like. The disorder frame and the neurodiversity frame arrive at the same place when the outcome data is followed honestly. The person does not need to be fixed. The gap does.

The rights shift described in this document is also a science shift. They are not separable. When the research population is defined by a protected characteristic rather than by who survived a clinical gatekeeping process, the sample changes. The research question changes. The findings change. And the findings travel.

Autism research conducted on clinically-gatekept populations has been studying a filtered sample and calling it a population. Every barrier in the diagnostic pipeline — the cost of assessment, the waiting list, the tools that couldn't reach women or non-speaking people or anyone whose presentation didn't match the 1943 sample — was also a filter on the research base. The people who didn't make it through the pipeline were not absent from the condition. They were absent from the data. Their absence shaped every finding — not as a limitation acknowledged in a methods section, but as a structural feature of what questions could even be asked.

Research must include community direction and must study clinical populations — not exclude them. Community-included research is required for lived experience, priority-setting, and identifying what the clinical sample missed. Clinically-grounded research is required for anything where diagnostic fidelity affects what the finding means — pharmacology, intervention efficacy, support need modelling. These are not in tension. The current conflation — using self-report community samples for clinical findings, or excluding community from research design entirely — produces bad science in both directions. The population defined by characteristic provides the broader base. The clinically confirmed sample provides the rigour where rigour is required. Both are needed. Neither replaces the other.

A population defined by characteristic rather than clinical certification contains the people the pipeline missed. It studies autistic people living in the world — in workplaces, in relationships, in education systems, in healthcare encounters — rather than in the clinical encounter where the assessment captures a snapshot of a person performing for an instrument never designed to see them. The question stops being what causes this deviation from normal and becomes what does this neurology need to function. The question stops being how do we reduce the symptoms and becomes what produces the symptoms — what environmental conditions, what support withdrawals, what mismatches between neurological architecture and built environment are generating the presentations the clinical system is currently managing as individual pathology. That shift produces immediately applicable findings. What sensory environments produce what neurological states. What support structures prevent crisis rather than arriving after it. What masking costs across a lifespan. The masking research alone, drawn from a population-level sample, will rewrite significant portions of the mental health literature.

The mental health crisis is in part an unidentified neurodevelopmental difference crisis. The people cycling through mental health systems on their fourth medication trial and their third diagnosis — BPD that doesn't resolve, depression that doesn't lift, anxiety that spikes in every environment the person enters — a significant proportion of them are autistic people whose upstream condition was never identified because the pipeline was never built to find them. Treating the downstream crisis without identifying the upstream condition is expensive, harmful, and produces the outcomes the system is currently documenting as treatment resistance. It is not treatment resistance. It is the wrong treatment. When the population is correctly defined, those people enter the research base. The mental health findings change because the sample changes. The intervention models change because the correct questions are finally being asked of the correct population. The cascade is not a hope. It is a methodological inevitability. Fix the sample. The science follows.

The resource crisis the restriction argument claims to address is real. The solution it proposes — narrowing the diagnostic category — addresses the crisis by reducing the entitled population. That is not a resource solution. It is an accounting trick. The people removed from entitlement do not become less expensive. They become more expensive — they move into mental health crisis services, emergency departments, homelessness infrastructure, justice systems, all of which cost vastly more per person per crisis than the support that would have prevented the crisis in the first place. The restriction argument has never produced a costing of what happens to the people it removes. That costing exists in the outcome data.

The support model that is both clinically correct and fiscally defensible is the individual plan — not because it is generous but because it is accurate. A population that is neurologically plural, with dynamic non-linear support needs shaped by environment, co-occurring conditions, cultural context, and intersecting identity cannot be efficiently served by a categorical funding model. Categorical models are built on the assumption that the category predicts the need. The neuroscience of autism confirms that the category does not predict the need. Presentation fluctuates with environment and support. Support needs are not linear across the spectrum. The same person can present profoundly differently depending on what scaffold is or is not in place. A categorical model applied to a non-linear dynamic population will always be wrong for a significant proportion of the people it serves — over-resourced in some cases, catastrophically under-resourced in others, with the under-served portion eventually hitting crisis at a cost that dwarfs the support that would have prevented it.

The individual plan resolves this as a methodological response to what the population actually is. Assess comprehensively and intersectionally — autism, co-occurring conditions, sensory profile, communication needs, environmental factors. Build a plan around what the person actually needs across domains. Attach that plan to the person, not to a diagnostic category or a funding threshold. Let it travel across systems — education, health, employment, housing — so that each encounter begins with the full picture. The GP as coordinator — appropriately positioned to provide general accommodation notes, hold the longitudinal picture, and coordinate specialist referrals — holds the intersectional picture across time, sees the pattern before it becomes a crisis, and coordinates referrals that the current system produces as siloed encounters each billing independently for a fragment of a picture none of them fully holds. The coordinator role and the formal diagnostic role are distinct clinical functions. Formal autism assessment requires specialist training, standardised instruments, developmental history across multiple contexts, and differential diagnosis of co-occurring conditions. Conflating the two is one of the system's current failures, not a feature of the model proposed here. The OT modifying the environment before the sensory overload produces the behavioural crisis that produces the psychiatric referral. The behavioural support understanding what the behaviour is communicating and addressing the communication need rather than extinguishing the signal while the underlying need remains. The psychiatrist working from the full neurological picture rather than as the first and only port of call for every autistic person in distress.

Reassessment across the lifespan is not optional. It is the clinical mechanism that keeps the individual plan accurate as the person ages, as co-occurring conditions emerge or change, as environments shift, as life transitions produce new support needs the original assessment could not have anticipated. A plan built at eight does not describe the person at twenty-two. Lifespan reassessment is how the plan stays honest. But reassessment is not the same clinical question as support withdrawal. The system currently treats them as identical. They are not. Reassessment asks: what does this person need now. Support withdrawal asks: can we remove what this person currently has. These questions require different evidence, different clinical justification, and different legal thresholds. Collapsing them — using reassessment as the mechanism for withdrawal — is the error the outcome data keeps documenting.

The evidence is specific and published. Studies following autistic children through early intervention found that a subset no longer met diagnostic criteria after receiving appropriate supports. The system read this as successful treatment — as evidence that the supports had worked and were therefore no longer needed. The supports were removed. Regression followed. The regression was documented as evidence of underlying severity or treatment failure rather than as the direct and predictable consequence of removing the scaffold that was producing the stability. The pattern is consistent enough to have a name even if the system has not given it one: support goes in, the child gets As, the As are read as evidence the support is no longer needed, the support comes out, the As stop, the system expresses surprise. The As were the support working. They were never the child not needing support. Reading the output as the baseline — as evidence of what the child can do without the support — is the methodological error expressed in a report card instead of a clinical file. It happens in classrooms, in funding panels, in reassessment meetings, across every system that measures output without accounting for what produced it.

The girl in a structured behavioural residential environment illustrates the same mechanism from a different angle. Inside a 24/7 structured environment she appeared to need level one support — manageable within the structure. The structure was doing enormous amounts of invisible work. The assessment measured her presentation inside that structure and read it as her baseline. It was not her baseline. It was the output of total environmental support being provided continuously and not counted as support need because it was ambient — because it was the environment itself. Remove the environment and the actual support need becomes visible. It was 24/7. It was always 24/7. The structure was providing it without anyone counting it as provision.

The safeguard must be stated explicitly: improved presentation under support is not independent justification for support withdrawal. Withdrawal requires its own clinical process — separately justified, separately documented. That process must establish what specifically has changed in the person's neurological picture independent of the support currently in place. It must include the person's own account of whether the support is doing work they cannot yet do without it. That is clinical information. A person who knows their stability is contingent on the scaffold currently in place is telling the clinician something the assessment cannot measure from the outside. Dismissing that account as lack of insight is not clinical judgment. It is the system protecting its budget by discrediting the person most qualified to describe their own support need. Reassess across the lifespan. Adjust the plan to what is actually needed now. But the plan cannot be used against the person. Stability is not evidence of independence from support. The cost of getting this wrong is not borne by the budget. It is borne by the person.

This model is also legally cleaner to uphold. An individual plan attached to a named person with specific documented obligations is enforceable in a way that categorical eligibility is not. The legal question is not whether someone meets a threshold that can be redrawn. It is whether the plan is being implemented. That is a narrow, specific, answerable question. The respondent cannot argue about category boundaries. They can only argue about whether they did what the plan required. The gatekeeping infrastructure that currently absorbs enormous administrative cost — reassessing diagnosis, re-adjudicating eligibility, managing appeals from people whose support was removed because their presentation improved under the support that was then removed — disappears. The plan travels. The support follows the person. The crisis that would have cost ten times as much to respond to does not happen.

The framework now being defended — the narrow model, the biological priority, the definitional authority of the research establishment over the community it studies — was built by excluding the community from the knowledge-production process. The CRPD's binding principle is not a procedural nicety. It is the recognition that research conducted on a population without that population directing it is not neutral science. It is extraction. The knowledge gets taken. The community pays the cost of what the knowledge gets wrong. And when the community challenges what the knowledge got wrong, it gets called corruption.

Research cannot inequitably discriminate. That is not an activist position. It is the legal framework that governs how states are permitted to conduct research on disabled populations. The IACC exists because Congress recognised that the community affected by autism research should have formal input into its direction. The documented failure of that input to change funding allocation is not a footnote to this argument. It is the argument.

The theory produced the results. Not the community. The community has been trying to correct those results for decades, through the correct channels, with the formal processes available to them. The record of that effort — and the record of what happened to it — is public, documented, and assembled above. Anyone who wants to challenge the argument has the sources. Anyone who wants to challenge the sources has the DOIs. The record is available to anyone who wants to read it before forming an opinion about what corrupted what.

The diagnostic restriction argument has a constituency. It is not the people it claims to protect. The autistic person who self-identified online, found community, and is navigating a workplace accommodation request will survive a tighter diagnostic system. They have language, access, and the capacity to fight. The autistic adult with high support needs whose residential funding, clinical plan, and legal protections are built around a formal classification will not survive it. They are already the hardest to correctly diagnose. They are already the most likely to be misclassified, overshadowed, routed toward punishment rather than support, and excluded from the research that is supposed to serve them. Narrowing the diagnostic criteria does not protect them. It removes the formal classification that is the only mechanism through which their support needs generate a legal obligation to meet them. The people who will pay the highest cost of diagnostic restriction are the people least able to show up to the policy debate to say so. The research on this is not contested. The outcome is predictable. The decision to proceed anyway is a choice.

There is a specific cruelty embedded in the high support needs erasure that the funding record alone does not capture. Studies document what practitioners and families often do not act on: many autistic people presumed to have minimal comprehension understand significantly more than the system has ever been designed to reveal. The child who knows the answer but has no recognised means to give it. Who watches the adults around them organise an entire life around a misreading. Whose parents grieve a child who is present. Whose teachers pitch instruction at an assumption rather than a person. Dawson et al.'s 2007 study — already cited in this document — showed a third of children labelled low-functioning scored in the average range when tested differently. The DSM-5 acknowledges IQ scores in autism are unstable in early childhood. The tool was measuring the framework's assumptions. The framework's assumptions determined who got what support. And the research that could have challenged that presumption earlier was not funded. The funding record is above.

The proposed solution — a separate "profound autism" subcategory — does not solve the problem it claims to address. It recreates it. The people who would be classified there already have multiple co-occurring conditions, each with its own diagnostic category, its own funding stream, its own research base. Intellectual disability. FASD. Epilepsy. Complex communication needs. Chromosomal conditions. If the support need is primarily captured by one of those frameworks, that framework exists and should be resourced. What the spectrum provides — and what a subcategory removes — is the recognition that the same neurological difference underlies all of it, and that autism-specific understanding of how that difference operates is clinically relevant regardless of support level. Splitting the spectrum at the high support end does not sharpen the tool. It breaks the framework that generates shared research, shared legal protections, and shared recognition of a common neurological reality. If the argument is that the high support population needs more research attention and better funded services — that argument is correct and this document supports it entirely. The answer is fund the research and build the services. Not create a new administrative enclosure and call it progress. Global developmental delay already exists. Intellectual disability already exists. What those categories cannot provide is autism-specific clinical understanding. That is what the spectrum provides. That is what a split removes.

The solution to the resource crisis is not diagnostic restriction. It is a tiered model in which most autistic people navigate identity, community, and protection from discrimination without requiring the formal clinical tier — and the formal clinical tier is available, properly resourced, and intersectionally complete for those who need it. Self-identification is valid. For those whose support needs, legal entitlements, or clinical picture require formal documentation, the pathway exists. The formal clinical tier encompasses not just mental health but hearing, vision, metabolic function, FASD, and co-occurring neurodevelopmental and psychiatric conditions — assembled in one place by a qualified clinician rather than across five separate referrals over ten years. A one-hour telehealth call does not meet that standard. The formal clinical tier produces an auditable record, identifies co-occurring conditions with their own obligations, rules out better explanatory frameworks where they exist, and generates a legally enforceable individual support plan. Neuroaffirming care as the standard throughout. Access to clinical supports does not indicate a higher degree of autism. It indicates that the intersectional profile requires those supports in the person's current environment. The proposed model does not describe an aspirational future state. It describes what competent intersectional clinical practice already produces when it happens — and what the current system's fragmentation consistently prevents. Build it properly. The gate takes care of itself.

Clinical documentation produced through the formal clinical tier is not an identity statement and it is not a badge. It is the legal mechanism that generates specific obligations — funded support plans, accommodation requirements, clinical treatment pathways. Most autistic people do not need it. For those who do, it is essential. The person whose residential funding, clinical plan, and legal protections are built around formal documentation cannot survive diagnostic restriction. They are already the hardest to correctly reach through the formal tier. They are already the most likely to be misclassified, overshadowed, routed toward punishment rather than support. Narrowing criteria does not protect them. It removes the documentation that is the only mechanism through which their support needs generate a legal obligation to meet them. And it does this in the name of a resource problem it does not actually solve — because the people removed from documentation do not disappear. They move into crisis systems that cost more.

The advocacy didn't corrupt the science. Community advocacy has repeatedly functioned as an external corrective mechanism, identifying gaps and failures that were later confirmed within the formal research record. Every major correction — refrigerator mothers, Wakefield, the life expectancy gap, the BPD misdiagnosis chain, Oliver McGowan — came from or was amplified by the community the establishment was trying to exclude from the knowledge production process. The community has functioned as an external correction mechanism for findings that the field's internal peer review processes did not identify or address. That is not a fringe position. That is the sequence of events documented above.

Every autistic person is part of the same neurological family — a spectrum as real and as varied as the spectrum of blue eyes, which also carries shared biological markers, shared vulnerabilities, shared things the medical system needs to know regardless of where on the spectrum a person sits. Autism does not need a cure. It needs competent clinical awareness of what comes with it — the medication sensitivities, the co-occurring conditions, the immune responses, the catatonia risk, the mental health intersections. Identity does not require documentation. Community does not require diagnosis. But the shared markers mean the medical system needs to know what it is working with — not to eliminate the neurology but to provide competent care to people who have it. When a child regresses the question is not which diagnostic category to place them in. It is: what is producing this regression. Environmental trigger. Support withdrawal. Co-occurring condition. Catatonia. Autoimmune response. The question that serves the child is what does this person need — not which side of a diagnostic boundary they belong on. The right question is the one the community has been asking for decades. The field funded the wrong one.

The governance standard that makes this work is already written. Person-centred means the autistic person's voice is the primary voice — not overridden by support level, not substituted by proxy, not managed by institutional convenience. Research cannot call itself autism research if autistic people are not directing it. Clinical practice cannot call itself person-centred if it substitutes professional judgment for personal choice. The CRPD supported decision-making framework is the binding standard. It already exists. It is already law. Apply it. The fight becomes unnecessary.

The ableist model assumed the people it was classifying could not connect, could not compare notes, could not build a counter-record. The isolation was load-bearing. The silo was the mechanism of control. The internet broke the silo — not by creating new autistic people or inflating the diagnosis, but by connecting the people who were always there and making the pattern recognition possible that institutional research was never going to fund. That assumption is now the thing that is collapsing. This document is part of that collapse.

The history of this field is not complicated. A narrow sample produced a narrow theory. A narrow theory produced a narrow research agenda. A narrow research agenda produced tools that couldn't reach the population they were supposed to serve, interventions that caused documented harm, and an outcome gap that the field's own data has been recording for decades without adequately explaining. The community identified each of these failures before the research apparatus did. The community was told it was corrupting the science. The corruption was already there. The community was the correction.

What this document has assembled is not a rebuttal. It is a record. The funding data is public. The outcome data is published. The legal frameworks are binding. The ethical guidelines are signed. The methodology that produces answers to the disorder framework's own unanswered questions is already written. The counter-example that proves the alternative works exists in multiple jurisdictions. None of this requires anyone's permission to be true. It requires only that the methodology be followed where it leads. That is already happening. The rights floor is expanding. The population is being defined. The research questions are changing. The political will follows the findings. The findings follow the methodology. The methodology is already obligated.

The autistic community has been waiting at the conclusion of this logic for decades. The science is still making its way there. The law is already arriving. The findings will follow. And when they do — when the population-level research directed by community priorities starts producing answers to the questions the disorder data raised and never funded — those findings will not stay inside autism research. They will move across every system where unidentified neurodevelopmental difference is currently being managed as something else entirely. The cascade is not a prediction. It is a methodological consequence of finally studying the right population with the right questions.

The theory produced the results. The community identified them. The methodology corrects them. What happens next is not contingent on any individual's willingness to arrive. It is contingent on whether the record exists and whether it can be found.

It exists. It can be found. That is what this document is for.