Diagnosis & Identity

Fad Diagnosis? What Kamp-Becker and Frith Overlook

"Someone who can hold a fluent conversation is not autistic." Uta Frith in an NZZ interview, 6 June 2026. The claim sounds clear. Neurologically it is wrong.

The fad-diagnosis claim has two prominent voices in German-speaking public discourse. Inge Kamp-Becker (Marburg) has framed "high-functioning autism" as a fad diagnosis for years and repeated the argument at the PTK Hessen conference in February 2026. Uta Frith, who formulated the Theory of Mind hypothesis with Alan Leslie and Simon Baron-Cohen in 1985, sharpens it in the NZZ: fluent speech rules out autism. Both claim to defend the term autism against inflationary self-identification. In practice they describe a behavioural phenotype and call it neurology.

What the critique gets right

TikTok content on autism is in parts inaccurate. Aragon-Guevara et al. (2023) documented that 41 percent of the most-viewed videos were inaccurate and 32 percent overgeneralised. A more recent analysis from the University of East Anglia (2026) confirms the picture. That is a real task for clinical communication. It is not evidence that the diagnosis numbers, which have risen for two decades, reflect a fashion. The public debate conflates the two - and that is where the argument tips over.

What is neurologically wrong

Frith's sentence assumes that observable social competence rules out autistic neurology. Four mechanisms contradict it directly.

Masking is the sustained top-down deployment of learned social scripts by the prefrontal cortex. A masking person processes two tasks in parallel in real time: content, and social performance. Hull et al. (2017) and Cassidy et al. (2018) document the cost - exhaustion, burnout, elevated suicide risk. From the outside it looks like competence. It is high effort against one's own processing mode.

Monotropism (Murray, Lesser & Lawson 2005) describes attention that focuses intensely on a single processing stream rather than distributing across many. It explains both deep special interests and exhaustion in socially layered contexts. It is not coupled to whether someone can speak.

Predictive coding (Sinha et al. 2014, Pellicano & Burr 2012) describes a different weighting between prediction and sensory signal. Autistic brains process deviation from expectation differently; this produces sensory sensitivity and a higher cognitive load when things are unpredictable. Independent of language.

The Double Empathy Problem (Milton 2012, Crompton et al. 2020) shows empirically that communication difficulties between autistic and non-autistic people are bidirectional. Autistic-to-autistic communication transfers information efficiently. The "social competence" Frith treats as an exclusion criterion is therefore an adaptation to the non-autistic majority - not a neurological absence.

Why diagnoses rise

The rise is well documented and has clear causes. Loomes, Hull & Mandy (2017) report a male-to-female ratio of roughly 3:1 at diagnosis, while actual prevalence is likely more even. Lai & Baron-Cohen (2015) describe the "lost generation" of late-identified adults. Mandell et al. (2009) document racial disparities in US diagnostic identification. This is not fashion. It is the delayed visibility of groups that the diagnostic grid - built on white boys - missed for decades.

Framing rising numbers as fashion erases this structural under-identification. It shifts the burden of proof: not the system that missed people, but the people now becoming visible have to justify themselves.

The following conversations are taken verbatim from the app, unedited.

This answer comes from Autistic Mirror. You can ask your own questions about your situation.

Frith's blind spot: the refuted Theory of Mind

In 1985, Uta Frith, together with Baron-Cohen and Leslie, proposed the Theory of Mind hypothesis: autism as a deficit in representing others' mental states. The hypothesis shaped four decades of research and clinical practice. It does not hold empirically. Gernsbacher and Yergeau (2019) summarise the replication record: ToM predictions fail to replicate, the tests often measure processing and language factors rather than mental attribution, and the Double Empathy model (Milton 2012) explains the observations better. When Frith says today that fluent speech excludes autism, she is leaning on a deficit frame whose empirical base has eroded.

Kamp-Becker's blind spot: phenotype as mechanism

Kamp-Becker argues conservatively from clinical instruments: only what ADOS and ADI-R measure counts as autism. The problem is that these instruments were validated on cohorts that do not adequately represent masking adults, women or late-identified people. They measure observable behaviour in a structured test situation. Masking is the very ability to perform neurotypically in a structured situation. An instrument that cannot detect masking cannot then prove the absence of autism when it fails to find it. It proves its own limit.

Two phenomena, one conflation

The fad-diagnosis claim works rhetorically only because it folds two very different observations into the same word. On one side are the viral checklist videos on TikTok and Instagram ("five signs you are autistic"). Aragon-Guevara, Castle, Sheridan and Vivanti (2023) systematically reviewed the most-viewed autism videos on TikTok: 41 percent were factually inaccurate, 32 percent overgeneralised, and a substantial share rested on personal anecdote without mechanistic grounding. A more recent analysis from the University of East Anglia (2026) confirms the picture. Such checklists are methodologically unusable, because they list visible behaviour without the underlying mechanism and so invite misattribution. That is a real problem of platform logic and the attention economy.

On the other side are the delayed diagnoses of groups overlooked for decades: women, AuDHD adults, late-identified maskers, people of colour. Three independent data lines (Fyfe et al. 2026, Lai & Baron-Cohen 2015, Mandell et al. 2009) point to the same pattern: not a sudden rise in prevalence, but structural under-identification that is now being caught up. These diagnoses are issued through clinical processes, not from a TikTok video.

Both observations are real, and they describe different mechanisms. Overgeneralising platform content is a knowledge-transfer problem that better clinical communication can address. Delayed diagnoses of overlooked groups are not a problem; they are a correction. Folding both under the label "fad diagnosis" places delayed diagnoses under the over-identification suspicion that belongs to the checklists, and erases the structural gap that makes these diagnoses necessary in the first place. Being overlooked is not the same as being over-diagnosed.

The actually open question

The productive question is not "is autism a fad diagnosis". It is: which diagnostic process reliably captures autistic neurology even when the person can hide it well? Asking that question opens the door to Hull, Cassidy, Milton, Crompton, Loomes and Mandell. Avoiding it by dismissing the visibility shift as fashion does not protect the diagnosis. It protects a diagnostic status quo whose limits are already empirically described.

The fad-diagnosis claim sounds like methodological care. In reality it amounts to renewed exclusion of groups overlooked for decades, who are only now finding an explanation for their thinking, feeling and acting. If the old picture was blurry, the picture has to be updated, not the late-included excluded again.

One bright spot

The debate is forcing a conversation about the limits of classical diagnostics that was not public for decades. It is bumpy. It is happening. For people who are currently identifying themselves or who were diagnosed late, the fad-diagnosis claim changes nothing about your neurology. It only changes how loudly you still have to hear the old frames before the discourse shifts.

Autistic Mirror explains autistic neurology individually, for your situation. For yourself, as a parent, or as a professional.