What Uta Frith proposes and what the studies show

Executive summary

This part presents as faithfully as possible the proposal formulated by Uta Frith in her editorial of 4 August 2026, without applying the theoretical framework of Autismology. Frith begins from the observation that the current category of autism spectrum disorder (ASD) has become extremely heterogeneous and that its diagnosed prevalence has increased sharply. She asks whether the term “autism” still covers a sufficiently coherent set to retain the same clinical and scientific value.

The core of her proposal concerns the difference between people diagnosed during childhood and people diagnosed later. She links this distinction to several recent findings: different developmental trajectories, polygenic architectures associated with age at diagnosis, variation in psychiatric co-occurring conditions, changes in the male/female ratio with age, and transformations in diagnostic practices and the cultural context.

The study by Zhang et al. (2025) is one of the main empirical supports: it shows that age at diagnosis is associated with different developmental trajectories and polygenic factors. The authors nevertheless specify that these differences are organized along a gradient rather than as discrete categories. The study by Fyfe et al. (2026) likewise shows a substantial catch-up in diagnoses among women with age, while recent literature on late diagnosis indicates that there is no single, consensual age threshold defining what counts as “late.”

Frith also discusses masking, compensation, self-identification, the influence of social media, contemporary uses of neurodiversity, and the possibility of false positives. Her stated aim is to call for greater diagnostic precision and to examine whether some subgroups currently brought together under the same diagnosis should be distinguished more clearly.

Key points

• Frith’s text is an argumentative editorial, not a new experimental study.

• The proposal places particular emphasis on differences between early and late diagnoses.

• The evidence invoked documents genuine developmental, genetic, and clinical heterogeneity.

• The cited studies do not all establish the same thresholds or the same interpretations of this heterogeneity.

• Part A presents these elements without deciding whether they justify a new diagnostic division; that question is reserved for Parts B and C.

Main source: Uta Frith (2026), Psychological Medicine

A.1 — The text under study and its status

The text examined is Uta Frith’s editorial published on 4 August 2026 in Psychological Medicine under the title “Autism spectrum disorder: has it lost its meaning and is it leading to misdiagnosis?”. It is an argumentative and synthetic essay, not a new experimental study. Frith brings together several recent studies and several cultural or diagnostic developments in order to question the coherence of the current category of “autism spectrum disorder” (ASD). https://doi.org/10.1017/S0033291726105376

Her starting point is twofold: first, the concept of autism has changed considerably since the first descriptions of the 1940s; second, diagnosed prevalence has increased substantially. She argues that the consequences of these transformations for research and clinical practice need to be examined. https://doi.org/10.1017/S0033291726105376

The article includes, among other things, discussion of changes in diagnostic criteria, the reality and unity of autism, the uneven increase in prevalence, the hypothesis of splitting the spectrum, cultural factors that have changed the concept of autism, the tension between social and medical models of disability, co-occurring conditions, and the search for greater diagnostic precision. https://doi.org/10.1017/S0033291726105376

A.2 — The evolution of the diagnostic category

Frith places the discussion in the history of the progressive broadening of criteria. Categories that were once separate were brought together within the notion of a spectrum, and inclusion thresholds evolved. For her, the advantage of a broad category—recognizing diverse presentations—also has a possible cost: the wider the set becomes, the more difficult it is to identify what all its members actually have in common. https://doi.org/10.1017/S0033291726105376

She therefore does not merely observe that diagnosed people differ. She asks whether current heterogeneity has reached a point at which the same term covers phenomena sufficiently far apart to reduce clinical and scientific precision. https://doi.org/10.1017/S0033291726105376

A.3 — The increase in prevalence and the question of the diagnostic threshold

Frith emphasizes the massive increase in diagnoses and considers several possible explanations. Her argument is not that every increase must correspond to error: she discusses improved recognition, changing criteria, cultural changes, and the possibility that the diagnostic threshold has been lowered. https://doi.org/10.1017/S0033291726105376

In her summary, among the factors that may have lowered this threshold she cites the desire for inclusion, avoidance of the risk of missing a diagnosis, increased reliance on subjective experience or self-description, and acceptance of masking as an explanation for low clinical visibility. She adds that the growing popularity of the concept of autism may encourage self-identification and identity-seeking. https://doi.org/10.1017/S0033291726105376

This part of the argument is important for what follows: Frith links the numerical evolution of the diagnosed population to a transformation of the concept itself, and not only to improved detection of an unchanged phenomenon. https://doi.org/10.1017/S0033291726105376

A.4 — The proposal to distinguish early and late diagnoses

The heart of the editorial concerns differences between people diagnosed during childhood and those diagnosed in adolescence or adulthood. Frith considers the differences observed between these groups to have become marked enough to justify the hypothesis of “two large subgroups.” In the reading made in the Autistan preparatory studies, she goes so far as to consider that people diagnosed after age 15 might belong to a distinct diagnostic category, or even to several categories. https://doi.org/10.1017/S0033291726105376

This proposal does not mean that she is already setting out a new operational classification. Rather, she opens a nosological question: could the timing of diagnosis signal trajectories sufficiently different to require a more precise subdivision of ASD? https://doi.org/10.1017/S0033291726105376

To this end, Frith brings together several observations: developmental differences, polygenic profiles, the frequency of certain psychiatric co-occurring conditions, changes in the male/female ratio with age at diagnosis, and cultural changes surrounding late diagnosis. https://doi.org/10.1017/S0033291726105376 ; https://doi.org/10.1038/s41586-025-09542-6 ; https://doi.org/10.1136/bmj-2025-084164

A.5 — The study by Zhang et al. (2025)

One of the most important empirical supports is the study by Zhang et al., published in Nature in 2025. Using longitudinal data from several birth cohorts and genetic analyses, the study reports two types of socio-emotional and behavioral trajectories associated with age at diagnosis. It also estimates that about 11% of the variance in age at diagnosis can be attributed to the common genetic variants studied. https://doi.org/10.1038/s41586-025-09542-6

The authors describe two autism polygenic factors that are only modestly correlated with each other (genetic correlation of about 0.38). One is more strongly associated with earlier diagnosis and lower social and communication abilities in early childhood; the other is more strongly associated with later diagnosis, socio-emotional or behavioral difficulties becoming more apparent in adolescence, and stronger genetic correlations with ADHD and some mental-health conditions. https://doi.org/10.1038/s41586-025-09542-6

Zhang and colleagues conclude that polygenic architecture and developmental trajectories vary according to age at diagnosis and that these findings may help explain part of autism’s heterogeneity. They also, however, present gradients of genetic correlations related to median age at diagnosis, which becomes an important point in the methodological examination in Part B. https://doi.org/10.1038/s41586-025-09542-6

A.6 — The case of girls and women

Frith also draws on changes in the male/female ratio according to age at diagnosis. The Swedish study by Fyfe et al., published in the BMJ in February 2026, concerns nearly 2.76 million people born between 1985 and 2020. It finds that the male/female ratio decreases as age at diagnosis increases and approaches equality much more closely by adulthood. https://doi.org/10.1136/bmj-2025-084164

The BMJ authors interpret these findings as a reason to investigate why girls and women receive their diagnosis later than boys and men. Frith uses the changing sex composition of late diagnoses within her broader argument about differences between populations diagnosed early and late. https://doi.org/10.1017/S0033291726105376 ; https://doi.org/10.1136/bmj-2025-084164

A.7 — Masking, compensation, and clinical visibility

Frith discusses masking at length. She is cautious about the idea that an absence of observable signs can be explained by invisible camouflage: when used without sufficiently testable criteria, she argues, this explanation risks making the diagnosis difficult to falsify. https://doi.org/10.1017/S0033291726105376

She distinguishes this notion from compensation, meaning the explicit learning of strategies that make it possible to resolve some difficulties in another way. This distinction is important in her reasoning because the concepts of masking and compensation are often invoked to explain why some people were not identified in childhood. https://doi.org/10.1017/S0033291726105376

The article therefore does not deny that strategies can exist; it asks how to document them without turning their supposed invisibility into automatic evidence for the diagnosis. https://doi.org/10.1017/S0033291726105376

A.8 — Subjective experience, identity, social media, and “looping”

Frith also gives space to cultural transformations of diagnosis. She discusses the growing value placed on accounts of lived experience, self-identification, online communities, the search for an explanatory identity, and the role of social media. She also discusses “looping” phenomena: the way a category can change how people understand and describe themselves, which can in turn feed back into the category itself. https://doi.org/10.1017/S0033291726105376

She considers the possibility of a broad form of social contagion: ways of describing or interpreting one’s behavior can circulate socially without necessarily implying simulation or deception. For Frith, such phenomena may nevertheless contribute to increasing the number of people who recognize themselves in the concept of autism and seek a diagnosis. https://doi.org/10.1017/S0033291726105376

Data from Neumann et al. on clinicians’ reports of an increase in self-diagnosis and desired diagnosis among young adults are among the elements discussed in the preparatory work for this publication. https://doi.org/10.1016/j.ijchp.2025.100661

A.9 — Neurodiversity and the tension between social and medical models

Frith acknowledges that the neurodiversity movement has helped reduce stigma and make the experiences of the people concerned heard. She nevertheless considers the language of neurodiversity and the social model to fit some late-diagnosed adults more readily than people diagnosed very early who have very substantial support needs. https://doi.org/10.1017/S0033291726105376

She sees a tension here: on the one hand, autism is presented as a difference or identity; on the other, a medical diagnosis continues to play a role in access to rights, care, or support. For her, this tension contributes to the difficulty of maintaining a single diagnostic category covering very different functional realities. https://doi.org/10.1017/S0033291726105376

A.10 — Co-occurring conditions and other diagnoses

Frith also examines the higher frequency of certain psychiatric difficulties or other co-occurring conditions among people diagnosed later. She asks whether part of what is currently grouped under ASD might be better described by other diagnostic categories or by several distinct dimensions. https://doi.org/10.1017/S0033291726105376

Earlier work indeed shows associations between certain co-occurring conditions and age at first autism diagnosis. This observation can support several interpretations, which will be examined in Part B. https://doi.org/10.1002/aur.2808 ; https://doi.org/10.1111/acps.13345

A.11 — “Late diagnosis”: a concept with no single scientific threshold

Recent literature shows that the expression “late diagnosis” does not refer to a uniform threshold. A preregistered systematic review of 420 articles found that only 34.7% of studies gave an explicit threshold; thresholds ranged from age 2 to 55, with a mean of 11.53 years, a median of 6.5 years, and concentrations around ages 3 and 18. This dispersion does not mean that age at diagnosis is uninformative, but it shows that “late” is partly a research convention dependent on context, sample age, and service organization. https://doi.org/10.1002/aur.3278

This point usefully complements Frith’s editorial: if a new diagnostic boundary is to be proposed on the basis of age at diagnosis, one must first show that the chosen division is robust across different thresholds and does not depend mainly on a methodological convention. https://doi.org/10.1017/S0033291726105376 ; https://doi.org/10.1002/aur.3278

A.12 — Girls and women: catch-up data fit into a broader literature

The Swedish BMJ study is not isolated. A systematic review of barriers to diagnosis in girls and young women identified, among other factors that can delay or prevent recognition, compensatory behaviors, other people’s perceptions, lack of information and resources, and clinical biases. https://doi.org/10.1007/s40489-020-00225-8

A more recent systematic review and meta-analysis likewise concludes that there are indications of underdiagnosis or misdiagnosis among women and that some phenotypic differences or camouflage may contribute to the diagnostic imbalance between sexes. These findings do not establish that every female-male difference is purely a detection artifact, but they strengthen the hypothesis that age at diagnosis also depends on how tools and clinical expectations identify female presentations. https://doi.org/10.1007/s11065-023-09630-2

Thus, the fact that a larger proportion of women is found among late diagnoses is compatible with several explanations. It may reflect genuine developmental differences; it may also reflect a history of poorer detection; and both mechanisms may coexist. The raw finding therefore does not by itself decide between these interpretations. https://doi.org/10.1136/bmj-2025-084164 ; https://doi.org/10.1007/s11065-023-09630-2 ; https://doi.org/10.1007/s40489-020-00225-8

A.13 — Co-occurring conditions: a robust difference, but not a unique interpretation

Mental-health differences associated with age at diagnosis are also documented beyond the editorial. In the SPARK sample studied by Jadav and Bal, 4,657 autistic adults with a professional diagnosis were compared according to age at diagnosis; adults diagnosed later reported more psychiatric conditions. https://doi.org/10.1002/aur.2808

In a Danish registry study of 16,126 autistic people, Rødgaard and colleagues observed that the rates of eleven co-occurring conditions varied significantly with age at first autism diagnosis and that this age contributed strongly to the clinical heterogeneity observed. https://doi.org/10.1111/acps.13345

These findings establish an important association between diagnostic trajectory and co-occurring conditions. However, by themselves they do not determine whether co-occurring conditions are a cause of late diagnosis, a consequence of prolonged trajectories without recognition, a factor that masks autism diagnostically, a source of misdiagnosis, or several of these mechanisms at once. That interpretive question belongs to Part B. https://doi.org/10.1002/aur.2808 ; https://doi.org/10.1111/acps.13345

A.14 — Toward greater diagnostic precision

Frith’s general conclusion is not simply “abolish the spectrum.” She calls for greater diagnostic precision. Her editorial proposes examining the reasons for the broadening of the concept and the practical consequences of a category that has become very broad. https://doi.org/10.1017/S0033291726105376

Her central thesis is that there are now marked differences between people diagnosed in childhood and those diagnosed in adolescence or adulthood, and that these differences justify explicit investigation of subgroups and possible alternative categories. https://doi.org/10.1017/S0033291726105376

The National Autistic Society responded on the day of publication by stressing that research on subtyping exists but, in its view, remains theoretical, with no established clinical or diagnostic value at present, and that a new subdivision could increase stigma or compromise access to support. https://www.autism.org.uk/what-we-do/news/our-response-to-professor-dame-uta-friths-new-paper This response is not part of Frith’s thesis, but it immediately situates the public controversy prompted by her editorial.

A.15 — What Part A establishes—and what it does not decide

At this stage, several facts can be distinguished: the diagnosed population is heterogeneous; late diagnoses have become more frequent; developmental, genetic, and clinical differences associated with age at diagnosis have been observed; the sex composition of diagnoses changes with age; and cultural transformations affect how autism is known, described, and sought. https://doi.org/10.1017/S0033291726105376 ; https://doi.org/10.1038/s41586-025-09542-6 ; https://doi.org/10.1136/bmj-2025-084164

By contrast, Part A does not decide whether these elements demonstrate several fundamental forms of autism, justify a new classification, or show that age at diagnosis is the best subdivision criterion. Those questions are precisely the subject of the logical and methodological examination in Part B, followed by analysis according to the Autismology framework in Part C.