Why the way science measures human variation matters — and what current evolutionary biology actually says about diversity.
The technical distinction
In ESF terms, there are two fundamentally different ways to model variation between people.
Parametric variation treats every person as positioned along a shared dimension — IQ, sociability, sensory threshold, executive function, attachment security. Everyone has the same architecture; what differs is the setting on the dial. Variation is deviation from a norm. The further from the mean, the more atypical, the more in need of correction.
Architectural variation treats different people as differently built. The wiring is different, not the dial settings. There is no single dimension along which everyone can be rated, because the relevant comparison is not “more or less of X” but “structurally different solutions to the same regulatory problem.” A neuroperipheral architecture and a neurosocial architecture are not high and low versions of the same system. They are different systems with different costs, different capacities, and different optimal conditions.
Most of contemporary autism research uses parametric framing — including most of what is called “the science.” This matters more than it sounds.
What parametric framing smuggles in
Once variation is modeled as deviation from a norm, four assumptions follow automatically, whether the researcher intends them or not.
First, there is a correct value — the norm itself, often statistical rather than functional.
Second, there are better and worse positions along the dimension, defined by distance from the norm.
Third, distance from the norm is read as severity — more deviation, more impairment.
Fourth, the proper goal of intervention is to move people closer to the norm. Less deviation. Less variability. Less variance in the population.
None of these assumptions are stated. They are built into the measurement apparatus. A researcher who would reject all four explicitly will still produce results shaped by all four implicitly, because the methods carry the logic forward whether the researcher is paying attention or not.
The slide into selection logic
If variation is deviation, and deviation is severity, and the goal is to reduce deviation, then the cleanest way to reduce deviation in the population is to reduce the number of people who deviate. This is selection logic. It does not require malice. It does not require Naziism. It does not even require explicit policy. It is what the parametric framework recommends to itself when followed to its conclusion.
This is not a hypothetical concern. The statistical methods most autism research uses were developed by Francis Galton specifically to advance his eugenic project — regression to the mean, correlation, the normal distribution as applied to human traits. The IQ testing tradition descends directly from Galton through Pearson, Spearman, and Burt. The diagnostic framework of the DSM treats deviation from norm as pathology by definition. Modern preimplantation genetic testing for autism-associated variants is an active discussion in reproductive medicine. The Sukhareva history reminds us that autism research and selection-era medicine were not separate enterprises; the diagnostic categories themselves came partly out of that context.
The point is not that researchers using parametric methods are eugenicists. Most would be horrified at the suggestion. The point is that the formal apparatus carries the logic regardless of who is using it. Disavowing the politics does not remove the structure. The slope is built into the measurement, and only explicit, sustained values work prevents the slide.
What evolutionary biology actually shows
Eugenic logic depends on an outdated model of evolution: competitive, hierarchical, with diversity as a temporary disorder selection will eventually clean up. That model has been wrong for at least fifty years.
Current evolutionary biology consistently finds that life is cooperative at every scale that matters. Mitochondria and chloroplasts are the descendants of free-living bacteria that entered into endosymbiosis (Margulis); eukaryotic life is a cooperative venture. Multicellular organisms are cooperatives of cells. The human body is a holobiont — a coordinated ecosystem of human cells and microbial partners whose regulatory functions cannot be cleanly separated (Cryan, Dinan, Mayer). Niche construction (Odling-Smee, Laland) shows that organisms actively shape their environments, which then shape them back; the gene-environment dichotomy was never accurate. Multilevel selection (Wilson, Sober) shows that selection operates simultaneously at the gene, individual, group, and ecosystem levels, with cooperation routinely outperforming competition at the higher levels. Tomasello’s work shows that distinctively human cognition is cooperative cognition; we cannot think the way we think alone.
Within this picture, diversity is not the residue of incomplete selection. It is the substrate selection works on. Populations without trait diversity cannot adapt — they are fragile in exactly the ways the portfolio effect literature describes (Schindler and colleagues on diversified salmon populations as the canonical example). Systems without sufficient internal variety cannot regulate complex environments — this is Ashby’s law of requisite variety, foundational to cybernetics and complexity science. Robustness comes from diversity, not despite it (Kitano).
The implication is direct. A species whose cognition is cooperative, whose bodies are holobionts, whose regulation depends on requisite variety, and whose ecosystems require diversity to remain stable cannot afford the parametric framing of human variation. Reducing variance reduces capacity. Pushing everyone toward the norm narrows the regulatory bandwidth of the population.
Neurodiversity as foundation, not exception
In the ESF framing, neurological variation is not a deviation to be reduced. It is the substrate on which human cooperation operates. Different neurotypes contribute different regulatory functions; a neuroperipheral architecture monitoring environmental pattern operates differently from a neurosocial architecture coordinating interpersonal coherence, and a functional human community needs both. This is what cooperative neurodiversity names. It is not a moral claim layered on top of biology. It is what the biology, read accurately, requires.
This reframes the question that animates so much current research. The question is not what causes autism or how can we identify it earlier or how can we reduce its prevalence. Those are parametric questions, and they carry the slide built in. The question is what regulatory functions does this architecture serve, what conditions allow it to operate coherently, and what does the population lose when those conditions fail. Those are architectural questions, and they open onto a different science entirely.
What changes when the foundation changes
Three things shift when biology is read from the cooperative-complexity side rather than the parametric-selection side.
Research questions change. Instead of asking which traits to reduce, the question becomes which regulatory architectures exist, what each one is for, and what conditions support coherent function across them.
Clinical orientation changes. Instead of pulling people toward the norm, the work becomes supporting the regulatory bandwidth and fit of each architecture in its own terms.
Population framing changes. Instead of treating diversity as a problem to be managed, the population’s diversity is recognized as its adaptive capacity — what makes it possible for the species to regulate the complex environments we now inhabit.
The science most autism researchers are looking for — rigorous, predictive, integrating biology and lived experience, accounting for heterogeneity, generating useful interventions — exists. It is not available from the parametric framework. It requires starting from the foundation that cooperative evolutionary biology and complexity science have already established.
Neurological diversity is not the problem the science is trying to solve. It is the precondition for the science being possible at all.
This is an ESF perspective from the Center for Adaptive Stress (CAS), an independent nonprofit conceptual systems research organization developing the Evolutionary Stress Framework as a complexity-science lens on neurodevelopmental variation, stress physiology, and regulatory health. ESF is a conceptual framework, not medical advice. For foundational reading, see Hogenkamp, Sanghavi, and Natri (2026), “Toward an Emergent Paradigm for Neurodiversity and Health,” Autism in Adulthood, DOI: 10.1177/25739581261433443. CAS publishes at evostress.blog and ndstress.org.
Suggested further reading: Margulis on endosymbiosis; West-Eberhard on developmental plasticity; Odling-Smee, Laland, and Feldman on niche construction; Wilson and Sober on multilevel selection; Tomasello on cooperative cognition; Ashby on requisite variety; Kitano on biological robustness; Schindler and colleagues on portfolio effects; Sheffer and colleagues on early-warning signals in complex systems.