The objection the archetype frame had to answer
The recent pillar piece on ADHD and autism as archetypes argued that autism is not a separate biological disorder but an archetypal inventory of cognitive mechanisms — 34 mechanisms latent in all humans, threshold-crossing into “pathology” when intensity exceeds what the externally-imposed demand structure tolerates. The archetypal framing locates what the diagnostic apparatus catches at the level of mechanism profile, not at the level of underlying biology. Different biological substrates can produce the same archetypal profile; the diagnostic category catches the profile, not the substrate.
The natural objection to that framing has always been biological. If autism is just a mechanism inventory, the argument runs, what about all the real neurobiology — the brain differences, the genetic variants, the structural and functional findings accumulated across decades of autism research? Surely those findings show that the diagnostic apparatus is catching something biologically specific, not a phenotypic profile that arises from multiple biological routes. The archetype frame, by this objection, sidesteps the biological reality of the condition.
A new study in Nature Neuroscience, published 26 May 2026, addresses this objection directly. It does so not by denying the biology, but by showing what the biology actually looks like.
What the study found
Pagani and colleagues, working under senior authors Alessandro Gozzi (Istituto Italiano di Tecnologia, Rovereto) and Adriana Di Martino (Child Mind Institute, New York, and co-founder of the Autism Brain Imaging Data Exchange), conducted a cross-species functional connectivity analysis to test whether autism’s well-documented phenotypic heterogeneity reflects distinct underlying biological substrates. The methodology combined functional magnetic resonance imaging from 20 distinct genetic mouse models of autism with brain scans from 940 humans with idiopathic autism and 1,036 neurotypical controls — a multimillion-pound, multi-site, multi-species design built specifically to test the biological-heterogeneity hypothesis.
The analysis identified two biologically dissociable subtypes that replicated across species. The hypoconnectivity subtype showed reduced functional communication between brain regions and was associated with synaptic dysfunction — gene expression analyses confirmed that the affected brain regions were enriched for synaptic genes. The hyperconnectivity subtype showed increased functional communication and was associated with transcriptional and immune-related alterations — affected regions were enriched for immune-related genes. The hyperconnectivity subtype scored moderately higher on standardised autism severity measures. The two subtypes accounted for approximately 25% of the autism cases examined.
That last figure is significant. The two subtypes together account for roughly a quarter of the autism sample. The remaining three quarters fell outside either subtype — meaning the actual biological heterogeneity within autism is not bimodal but multimodal, with at least two reproducible substrate configurations now identified and likely several more sitting beneath the same phenotypic category, waiting for the right analysis to find them.
The substrate varies, the archetype holds
The architectural read of this finding is straightforward. Two qualitatively different biological substrates — one arising through synaptic dysfunction, the other through transcriptional and immune-related alterations — both produce something the diagnostic apparatus catches as autism. The biology varies. The phenotype the diagnostic apparatus catches stays stable. That pattern is empirically incompatible with the unitary-biology model of autism, and it is empirically consistent with the archetype frame.
If autism were a single biological condition, this finding should not have been possible. A unitary biological disorder would produce a unitary biological signature — a single connectivity profile, a single pathway, a single biological mechanism producing the diagnostic phenotype. What Pagani and colleagues found is not that. It is two qualitatively different biological substrates converging on the same archetypal phenotype, validated across species, with the explicit acknowledgement in the paper’s own framing that “direct evidence supporting” the link between autism’s phenotypic heterogeneity and underlying pathobiological variation had been “lacking” until this work demonstrated it.
This is precisely what the archetype frame predicts. The archetypal inventory of 34 cognitive mechanisms that constitutes the autism archetype is what the diagnostic apparatus catches at threshold-crossing intensity. The mechanism profile is the level at which “autism” is a meaningful category. The biology beneath it can be configured in multiple ways — through synaptic dysfunction in some individuals, through immune-related transcriptional alterations in others, and through the unknown configurations sitting in the three quarters of cases the study could not yet sort into the two identified subtypes — and still produce the same archetypal profile the diagnostic apparatus catches. Multiple biological routes converging on a stable phenotypic archetype is exactly what mechanism-level architecture looks like in real biology. It is not a complication of the architectural framing. It is its empirical confirmation.
The implication for autism research is structural. The decades-long search for the biology of autism — the gene, the brain region, the connectivity pattern that defines the condition — has been searching at the wrong layer. The biology is real and identifiable, but it is heterogeneous in ways that the unitary-biology model cannot accommodate. The phenotype is what is being caught. The biology is what underlies it, in different configurations for different individuals.
The remaining 75% of cases the study could not yet assign to either subtype is itself the strongest evidence that more biological heterogeneity is coming. Future analyses will likely identify additional configurations, all of which will sit beneath the same archetypal mechanism inventory. The pattern will keep replicating: the substrate varies; the archetype holds.
What this does and doesn't change
What this changes: the unitary-biology model of autism is no longer empirically defensible. Researchers and clinicians who continue to search for the single biological signature that defines autism are looking for something the empirical data, as it now stands, indicates does not exist. The biology of autism is plural, biologically dissociable, and likely far more heterogeneous than even this study’s two-subtype finding suggests. Treatment approaches built on the assumption of single-biology autism — and there are many — will increasingly find themselves operating on shaky empirical ground.
What this does not change: the diagnostic category remains useful, because what it is catching — the archetypal mechanism profile — is real and stable across biological subtypes. The lived experience of autistic people remains exactly what they describe it to be; the biological heterogeneity beneath does not alter the phenotypic reality at the mechanism level. The archetype frame remains intact, because the mechanism inventory it identifies is precisely what multiple biological substrates appear to converge on. And the structural critique this site has been building — that the diagnostic apparatus catches mechanism profiles at threshold-crossing intensity against an externally-imposed demand structure — remains unchanged. The substrate-level evidence simply confirms that the architectural framing was located at the right layer all along.
The field’s next move, if it follows the empirical evidence rather than the inherited unitary-biology assumption, is to recognise the archetype layer explicitly. The diagnostic categories called ADHD and autism are catching mechanism inventories. The biological substrates beneath those inventories are plural and structurally heterogeneous. The architectural framing the pillar piece set out a week ago is now sitting on top of the strongest empirical anchor it has yet had.
Citations
Pagani, M., Gozzi, A., Di Martino, A. et al. (2026) — Autism subtypes identified using cross-species functional connectivity analyses
