We disagree that the autism spectrum has lost its meaning, that cultural factors have driven diagnostic inflation, or that differences between early- and late-diagnosed individuals now justify division into separate categories (Frith, Reference Frith2026). These claims conflate several distinct problems and risk real-world harm.

Autistic Doctors International (ADI) numbers over 1,500 autistic medical doctors worldwide, demonstrating that autism is not incompatible with a medical career. Some were diagnosed as young as 3 years (Shaw et al., Reference Shaw, Fossi, Carravallah, Rabenstein, Ross and Doherty2023), meeting Frith’s criteria for the core group, yet by their professional profile are indistinguishable from colleagues she places in the other. Her taxonomy cannot accommodate this.

Some had childhoods where differences were conspicuous; others built careers of apparently effortless competence while quietly managing demands their neurotypical colleagues never had to consider. Yet across differing profiles they find commonalities – sensory differences, pattern recognition, and the exhaustion of life among a non-autistic majority – across what Frith frames as an unbridgeable divide.

That common phenomenology extends further still. A published exchange between one of us and a nonspeaking autistic undergraduate who communicates by independent typing found the experience of being perceived as a fraud spontaneously recognized across profiles Frith’s framework would separate (Doherty, Reference Doherty2026a). The extent of this shared phenomenology deserves rigorous investigation, not premature taxonomic foreclosure.

Frith insists on ‘disorder’, presented as scientific precision. This is a value-laden choice with consequences. An ADI membership survey found members viewing autism as a disorder were almost three times more likely to have attempted suicide than those who did not (Shaw et al., Reference Shaw, Fossi, Carravallah, Rabenstein, Ross and Doherty2023). Autistic shame – the internalized sense of defectiveness arising from lifelong exposure to neuronormative expectations, where difference is misread as deficit – is a barrier to disclosure among autistic psychiatrists (Doherty, Reference Doherty2026b).

This matters beyond the individual. Where shame impacts self-recognition or disclosure – including among researchers and clinicians – it shapes science: the questions asked, the experiences centered, the expertise recognized. How we name, frame, and shame these difficulties has consequences reaching well beyond diagnostic manuals, into consulting rooms, classrooms, and workplaces where autistic people spend our lives (Doherty et al., Reference Doherty, Rebowska, Grosjean, Kinnear, Dakin, Porter, Sadiq, Greeter, Henderson, Wijetilleka, Hill, Rabenstein and Mayer2026).

What Frith characterizes as loosening of criteria is the correction of a systematic bias. For decades, diagnostic criteria reflected white male presentations, often with co-occurring intellectual disability – thus excluding most autistic people. The category was never a pure one later contaminated; it was always incomplete. Late diagnoses among women and people without intellectual disability are evidence of historical diagnostic failure: people rendered invisible by narrow criteria, not invented by wider ones. Many spent decades misdiagnosed with depression, bipolar disorder, schizophrenia, or personality disorders, treated with pharmacotherapy that never addressed the underlying neurodevelopmental reality.

The most substantive point Frith offers rests on the finding that polygenic profiles differ between early- and late-diagnosed autistic people (Zhang et al., Reference Zhang, Grove, Gu, Buus, Nielsen, Neufeld, Koko, Malawsky, Wade, Verhoef, Gui, Hegemann, Geschwind, Wray, Havdahl, Ronald, St Pourcain, Robinson, Bourgeron and Consortium2025), the later-diagnosed group showing stronger genetic correlation with ADHD, major depressive disorder, and PTSD. She reads this as misclassification, an inference requiring scrutiny. Polygenic correlations do not establish the direction or mechanism of association: they are also consistent with these conditions being downstream consequences of unrecognized autism. Frith rejects the concept of masking, yet it is robustly associated with elevated anxiety, burnout, depression, and suicidality, and the genetic profiles of masking and compensation have not been examined. A large-scale integrative study found elevated depression and ADHD polygenic scores across autistic people without intellectual disability generally, not only late-diagnosed subgroups (Fleury et al., Reference Fleury, Mougin, Leblond, Cliquet, Mathieu, Oakley, Maruani, de Saint Agathe, Korkmaz, Holt, Vitrac, Amsellem, Batut, Botton-Amiot, Rolland, Barthome, Bonicel, Beggiato, Lemière and Bourgeron2025). Zhang’s findings cannot bear the weight of misclassification, Frith asserts.

Frith characterizes the later-diagnosed group primarily through social anxiety – a significant empirical reduction. Autism is not anxiety. The sustained effort of performing neurotypicality in a world not built for autistic cognition leads to anxiety and to unexplained difference internalized as personal failing. Calling masking unfalsifiable does not survive the measurement literature: it is operationalized in validated instruments with consequences documented in population-based research (Cook, Hull, Crane, & Mandy, Reference Cook, Hull, Crane and Mandy2021). Nor does the dismissal that ‘we could say we are all masking, all the time’ – when masking is your unexamined baseline, the concept becomes invisible because one has no reference point outside personal experience. Autistic professionals describe recognizing, often only after decades, that what felt like ordinary social effort was qualitatively different (Doherty, Chown, Martin, & Shaw, Reference Doherty, Chown, Martin and Shaw2024). The absence of visible impairment in a well-accommodated environment indicates successful accommodation, not the absence of autism; remove the scaffold and autistic needs rapidly reappear.

Frith’s division rests, finally, on age of diagnosis – not a neurobiological variable but a function of access barriers, gender, race, socioeconomic status, and clinician awareness. Building such a taxonomy would institutionalize existing inequities in who gets recognized, when, and by whom.

The pressures on educational and health services created by the rise in autism diagnoses are real and require serious policy attention. But they are policy problems, not scientific ones, and allowing resource pressures to drive nosological revision would be a category error. If the response is to reframe some autistic people as not quite autistic enough, the likely outcome is not better support for those with greatest need, but the withdrawal of recognition from a population already facing substantial barriers to identification.

Frith is right that autism is heterogeneous. But heterogeneity is here being used to fracture the category – to decide that some autistic people are sufficiently autistic and others are not – rather than as a prompt for deeper investigation. Its consequences fall hardest on those it purports to protect. Shared autistic experience – articulated by those who can communicate, and not only in spoken words – illuminates sensory processing differences, the harms of compliance-based interventions, autistic burnout, and the potential of augmentative and alternative communication. Separating the community into groups whose insights cannot inform one another deprives those with the highest support needs of the knowledge most likely to improve their outcomes (Doherty, Reference Doherty2026b). What is required is sustained, co-produced investigation of shared and differing phenomenology across autistic profiles – research that does not presuppose its conclusions.

The spectrum has not collapsed; our understanding remains incomplete. Scientific constructs that accommodate new evidence, expand to include previously invisible populations, and resist premature closure are working, not failing. What Frith describes as conceptual collapse is the belated recognition of people who were always there.