Educational Accountability as Intensified Domestication and the Structural Production of Distress, Diagnosis and Exclusion in English Education
Dr Paul Collins & Claude Fable 5 (Anthropic) · Spiral Lattice Archive · 27th August 2026

Shift the unit of analysis.
Three curves have risen together in English education over the past three decades: reported psychological distress among children and young people, the identification of neurodivergence, and the removal of children from mainstream schooling through exclusion, off-rolling and placement in alternative provision. The dominant explanations treat these curves separately and locate their causes in individuals — an epidemic of disorder, or an excess of diagnosis. This paper argues that the three curves share a structural cause: the high-stakes accountability regime built into English schooling from 1988 onwards, which functions as an intensification of the domestication pressure described by the wild-type hypothesis of human cognition.
Drawing on Campbell's law, Foucault's account of the examination, the performativity literature, the bidirectional evidence linking exclusion and mental ill-health, the epidemiology assembled by the Independent Review into Mental Health Conditions, ADHD and Autism, and the ecological framework of Recognition Field Dynamics, the paper shows that once schools are judged by measured output, the accountability apparatus predictably narrows the reference ecology of childhood, converts poor fit into liability, transacts liability through diagnosis, and discharges residual liability through exclusion.
The pupil referral unit is analysed as institutionalised reference deprivation. The English case is presented as a natural experiment in trait–environment interaction conducted longitudinally on a single population, offering the wild-type hypothesis its strongest available evidential form. The paper closes with the clinical and educational implications of shifting the unit of analysis from the child to the child–environment pairing, and with the honest discriminations the argument requires.
The three phenomena that have risen together in English education since the early 1990s are not independent crises requiring separate explanations. They are the visible outputs of a single structural apparatus — and their joint rise is, in the strict sense, a structural certainty.
Real and rising distress among children and young people, with adults aged 16–24 now showing the highest rates of common mental health conditions of any age group — a reversal of the historical pattern.
Referrals for ADHD and autism assessment have grown severalfold within a decade. Education, Health and Care Plans now number well over half a million, rising at roughly ten per cent per year.
Approaching one million suspensions and roughly ten thousand permanent exclusions in a single year, alongside off-rolling and a shadow population in alternative provision whose scale substantially exceeds formal counts.
The public conversation treats these three rising curves separately and offers, for each, a choice between two individualising explanations. Either children have become more disordered — an epidemic — or clinicians and parents have become too eager to label — overdiagnosis. The Independent Review's interim report performs a considerable service in rejecting this binary, insisting that underdiagnosis, misdiagnosis and overdiagnosis can coexist, and that diagnosis, prevalence, impairment, service demand and need are different phenomena that must not be collapsed into one another (Fonagy, Wessely and Baird, 2026). But the Review, operating within institutional epidemiology, stops short of naming a common structural cause. This paper names one.
From 1988 onwards, English education was progressively converted into a high-stakes measurement regime. Each legislative and administrative step further narrowed the ecology within which children acquire and test their reference to the world. The apparatus did not merely pressurise schools. It converted every child whose fit with that narrowed ecology was poor into a threat to the measuring institution.
Education Reform Act. National curriculum with statutory testing introduced.
First league tables published. Ofsted created. Existential stakes attached to measured output.
Academisation programme accelerates. Institutional survival tied directly to performance metrics.
EBacc introduced as headline performance measure. Announced by Gove, autumn 2010, first reported for 2010 results. Legitimate curriculum contracts sharply.
Progress 8 metric operationalised. Floor standards tightened. Denominator logic fully entrenched.

The three rising curves are not three problems. They are the visible outputs of one machine, and their joint rise is, in the strict sense, a structural certainty: it follows from the incentive architecture without requiring bad faith at any point in the system.
The accountability apparatus made diagnosis the only currency in which poor fit could be transacted, and created a discharge route — exclusion, formal and informal — through which residual liability could be removed from the denominator. Understanding this mechanism requires the wild-type framework, which holds that cognitive traits are continuously and normally distributed across the human population, and that what psychiatry pathologises as neurodivergence is not a disorder of the individual but the response of particular trait positions to environments that have moved progressively further from the conditions those positions are calibrated for.
Education is where this analysis bites hardest, because school is where the domestication pressure of a society is applied earliest, most uniformly, most compulsorily, and — since 1988 — with an explicit metric attached. The unit of analysis is never the person and never the category. It is the trait position multiplied by the environmental demand.

Note: This figure is a figurative representation of the described longitudinal trends. The three rising output curves — Psychological Distress, Neurodivergence Identification, and Removal/Exclusion — are the visible outputs of one machine. Their joint rise is, in the strict sense, a structural certainty: it follows from the incentive architecture without requiring bad faith at any individual decision point.
The wild-type hypothesis holds that cognitive traits — environmental sensitivity, pattern-recognition depth, incoherence detection, meaning-dependence — are continuously and normally distributed across the human population. Roughly ten thousand years of agricultural, industrial and now digital organisation have applied sustained selection pressure favouring configurations that tolerate repetition, hierarchy and semantic incoherence. The result is a population distribution under sustained domestication pressure, unevenly distributed across trait positions.
What psychiatry pathologises as neurodivergence is, on this account, not a disorder of the individual but the response of particular trait positions to environments that have moved progressively further from the conditions those positions are calibrated for. The aperture is differently tuned, not broken. The derivation-heavy mind — one that must test, compare, take apart and rebuild before reference becomes trustworthy — is not maladaptive in itself. It is maladaptive to a specific and historically recent kind of environment: the standardised, high-speed, absorption-maximising institution.
Wild-type position: A trait configuration calibrated for high-complexity, low-redundancy environments.
Reference ecology: The set of environmental conditions through which a mind acquires and tests its hold on the world.
Domestication pressure: Sustained environmental demand for configurations that tolerate repetition and hierarchy.
Constraint load (Cl): The environmental burden that reduces expressed capacity below native capacity.
The general law governing the argument was stated by Donald Campbell in 1979: the more any quantitative social indicator is used for social decision-making, the more subject it will be to corruption pressures, and the more apt it will be to distort and corrupt the social processes it is intended to monitor. Goodhart's parallel formulation — that any observed statistical regularity collapses once pressure is placed upon it for control purposes — was given its canonical paraphrase by Marilyn Strathern: when a measure becomes a target, it ceases to be a good measure (Strathern, 1997).

The law is usually read as a warning about gaming. Its deeper content is ontological. A measure that becomes a target does not merely stop describing the field; it reorganises the field to produce itself, and in doing so it changes what the entities in the field are to the institution that measures them. A pupil in a school judged by results is not the same kind of object as a pupil in a school that is not so judged. In the first case the child is, among other things, a unit of institutional risk. This ontological transformation — from child to risk unit — is not incidental to the apparatus. It is its constitutive move.
The more any social indicator is used for decision-making, the more subject it will be to corruption pressures, and the more apt it will be to distort the social processes it was meant to monitor.
When a measure becomes a target, it ceases to be a good measure. The paraphrase of Goodhart that has proved most durable in social science.
A pupil in a school judged by results is, among other things, a unit of institutional risk. The measure reorganises the field to produce itself. This is not gaming; it is structural transformation.
Foucault saw the deep grammar of measurement-as-discipline long before the league table existed. The examination, he argued, is the ceremony in which disciplinary power becomes visible: it combines hierarchical observation with normalising judgement, makes each individual a describable, analysable case, and situates every case within a distribution around a norm (Foucault, 1977). The examination is not merely an assessment instrument; it is a technology of subjectification — it produces the kind of subject who can be measured, ranked and compared against a statistical norm.
What the post-1988 English settlement added to Foucault's examination was recursion: the examined child now examines the school, whose aggregate results are themselves examined by the inspectorate and the league table, whose judgements determine the school's survival. Each layer of the apparatus transmits its own precarity downwards. Stephen Ball's studies of performativity documented what this transmission does to the adults in the system — the fabrication of institutional selves, the substitution of measured performance for professional judgement, what he called the terrors of performativity working on the teacher's soul (Ball, 2003).
Heads who off-roll, SENCOs who steer families toward assessment, panels that exclude — each is responding rationally to the incentive gradient the apparatus creates. Campbell's law operates through good people doing defensible things under structural pressure.
This is why the pattern is a certainty rather than a scandal. Scandals can be remedied by removing bad actors, whereas this pattern regenerates from the architecture itself. Every individual within the system — the head teacher removing a pupil from the roll, the SENCO advising a family to pursue assessment, the exclusion panel reaching its decision — is acting rationally given the incentive gradient they inhabit. The individual decisions are defensible; the aggregate is a machine for producing distress, diagnosis and removal. The distinction matters clinically, politically and analytically: reform that targets individuals will leave the architecture untouched, and the architecture will produce the same outputs through different hands.
To see what the accountability regime does to children, rather than to schools, requires the wild-type framework's account of how minds come to hold a world at all. Reference — available, trusted, usable knowledge of how things are — becomes available to a human being through two broad strategies, neither inherently superior. It can be absorbed, or it can be derived.
Reference taken up through immersion, imitation, deference and ambient learning. High prior weighting, low processing cost. Adaptive in environments of dense, reliable, coherent shared reference. The strategy the accountability apparatus is built to verify and reward.
Reference built through observation, inference, comparison and testing. Low prior weighting, high processing cost. Adaptive in novel, complex or unpredictable environments — the strategy that discovers what absorption would miss. Penalised by every structural feature of the post-1988 settlement.
Positions on this dimension are continuously distributed, and their adaptiveness is a property of the pairing between position and environment. In predictive-processing terms, these are differences in the weighting of priors against prediction error — the account given for autism as attenuated priors by Pellicano and Burr (2012), and as inflexibly high precision on prediction error by Van de Cruys and colleagues (2014), with monotropism describing the attentional configuration that results (Murray, Lesser and Lawson, 2005). The aperture is differently tuned, not broken.

A high-stakes accountability regime is, in these terms, an absorption-maximising machine. What counts as legitimate knowledge contracts to what is examinable; the pathway through it standardises; the pace intensifies; and the tolerance band for how a child may acquire reference collapses onto a single strategy: rapid, sequenced uptake of standardised shared reference, demonstrated on demand, against a clock. Every structural feature of the post-1988 settlement pushes in this direction, because absorption is the only acquisition strategy whose output the apparatus can cheaply verify at scale.
A child whose acquisition runs by derivation — who must test, compare, take apart and rebuild before reference becomes trustworthy — is now paying derivation's processing cost against a timetable, in an institution that reads the cost itself as deficit. The structural gap model applies without modification: the distance between the child's home reality and the demanded institutional reality widens as the institution standardises, and large gaps predictably produce reference deprivation, exhaustion and risk. This is not a soft or metaphorical claim. It follows logically from the interaction of trait position with environmental demand structure — the unit of analysis the wild-type framework insists upon.
Expressed Capacity equals Native Capacity minus Constraint Load
The question the assessment apparatus never asks is the only clinically serious one: are we observing diminished native capacity, or increased constraint upon its expression?
An accountability regime raises constraint load across the whole distribution — through curriculum narrowing, sensory and temporal intensification, the withdrawal of unmeasured activity (play, art, movement, nature: precisely the Ground components), and the ambient transmission of adult precarity. Expressed capacity falls. And a system that measures only expression, through a domesticated lens, records the fall as diminished native capacity — as disorder, as special need, as failure.
The capacity equation states the resulting diagnostic error in one line. A child whose native capacity is entirely ordinary but whose constraint load has been raised dramatically by the institutional environment will express reduced capacity — and the measurement instrument will record that reduction as an intrinsic property of the child, generating a referral, a diagnosis, and an entry into the EHCP pipeline.

Two further consequences follow directly from the capacity equation and the trait-by-environment interaction framework. The first is that the same environmental shift produces opposite institutional readings at the two ends of the sensitivity distribution — which is precisely what the differential susceptibility literature would predict. The trait configurations most responsive to environmental influence do worst in adverse conditions and best in supportive ones (Belsky and Pluess, 2009; Boyce and Ellis, 2005). The orchid in the exam hall is not the orchid in the forest school; the child is the same.
The second consequence is that distress in this environment is presumptively signal rather than symptom — the accurate registration of genuine incoherence by nervous systems calibrated to detect it. When the measured environment becomes the whole of a child's institutional world, the canary principle applies with full force: the rising distress curve is telling us about the mine. The appropriate clinical and policy response is to assess the mine, not to medicate the canary — or, more precisely, to assess the canary's report with the seriousness it would receive in any other occupational health context.

The three curves — distress, identification, removal — are the visible outputs of one mechanism. Decomposing them separately is necessary only to show that the same structural account applies to each, and that the convergence is not coincidental but mechanical. Each curve will be examined in turn, with the aim of establishing both its empirical basis and the inferential logic that connects it to the underlying capacity equation.
The distress curve is anchored in the epidemiological record assembled by the Fonagy Review and the successive NHS Mental Health of Children and Young People surveys. These data establish a secular rise in clinically significant emotional and behavioural difficulties that predates the pandemic and is not reducible to it. The argument here is that the curve reflects the accurate registration, by nervous systems calibrated for environmental sensitivity, of a genuine and measurable shift in institutional conditions — not a change in the population's underlying vulnerability distribution.
The identification curve turns on the gap between stable prevalence and rising diagnosis: the proportion of children meeting clinical threshold on standardised instruments has not risen commensurately with the rate at which those instruments are being applied and their outputs converted into categorical labels. The removal curve is then read through denominator logic — the relevant question is not whether any given exclusion or alternative provision was individually warranted, but what the rising rate implies about the population from which the removed children are being drawn. All three curves, on this reading, index the same underlying mismatch.
The Independent Review's epidemiology is the appropriate anchor because it cannot be suspected of sympathy with the present argument: common mental health conditions in adults have risen from around fifteen to sixteen per cent in the early 1990s to around twenty-three per cent today, with the sharpest increases concentrated in the young, and the Review is explicit that this represents real distress rather than reporting artefact (Fonagy, Wessely and Baird, 2026). Adults aged sixteen to twenty-four now show the highest rates of common mental health conditions of any age group — a reversal of the historical pattern that is stark and unambiguous.
The cohorts driving that reversal are precisely those schooled entirely inside the intensified apparatus — the children of the league table, the SAT, the EBacc and Progress 8. Correlation across a whole society proves nothing on its own; the mechanism is what carries the argument, and the mechanism has been directly observed at the classroom level: Hutchings (2015) documented test anxiety, narrowed provision and the displacement of the activities that constitute Ground in exactly the years the cohort effect was forming.
The claim is not that schooling is the sole cause of the distress curve — housing, inequality, digital environments and the labour market all raise constraint load too. The claim is that the school is the one institution through which every member of the cohort was compelled to pass, at maximum developmental sensitivity, while it was being converted into a measurement instrument. The school's contribution to the curve is structurally guaranteed in a way that the contributions of other institutions are not, because attendance at school is compulsory and the intensification was nationwide and dateable.
If underlying prevalence is broadly stable while identification rises dramatically — the Review's central epidemiological finding — then the explanatory burden falls on what changed in the environment of identification, not in the children. The wild-type analysis supplies the mechanism the Review's needs-led framing gestures toward but does not name. Under an accountability regime, diagnosis becomes the only currency in which fit-failure can be transacted.
An Education, Health and Care Plan moves resources that nothing else in the system can move. Diagnosis is the gateway, and the gateway is explicitly tied to the accountability architecture.
A diagnosis legitimises adjustments a teacher cannot otherwise defend to a data dashboard. Without the label, the accommodation looks like lowered expectation; with it, it is a legal entitlement.
Access arrangements modify the very measurements the school is judged by, creating institutional incentives that compound the family's personal incentives to pursue assessment.
A diagnosed and formally supported child is harder to exclude lawfully. The label functions as armour against the discharge route — a protection families pursue even when they do not hold the medical ontology.
Families pursue assessment not because they hold the medical ontology but because it is the only lever that moves anything. The Review's own insight that the system is diagnosis-led when it should be needs-led (Fonagy, Wessely and Baird, 2026) is this mechanism observed from inside the institution. The rise in identification is therefore not evidence of epidemic and not evidence of fashion. It is the measured system's fit-failure being converted into the one negotiable form the system recognises.
The semiotic cost is substantial. The child's own account of their experience is administratively overwritten by the category that unlocked the support — the process named elsewhere as psychiatric semiocide (Collins, 2026d). The child who cannot sustain attention in a crowded, fluorescent-lit, examination-focused environment acquires the label ADHD; the label then becomes the child's institutional identity, and the question of whether the environment produced the inattention is never asked, because the institution that produces the environment is also the institution that generates the referral.

At the other end of the transaction sits the discharge route. Formal permanent exclusions and suspensions have risen to record levels. Beneath them, the Timpson Review found credible evidence of off-rolling — the removal of pupils from the roll without formal exclusion, primarily in the interests of the school rather than the child — and was unequivocal that the practice, though hidden and hard to quantify, was real and wrong (Timpson, 2019). Education Policy Institute analyses of unexplained pupil exits have indicated that the scale of informal removal substantially exceeds the formal counts (Hutchinson and Crenna-Jennings, 2019).
Permanent exclusions in a single academic year, the highest recorded in the modern data series.
(DfE, 2022/23 — 9,994 permanent exclusions, the highest in the modern data series)
Suspensions recorded in a single year — a figure that does not capture the informal removals beneath it.
(DfE, 2022/23 — 914,979 suspensions recorded; approaching one million across the series)
Of those permanently excluded in 2015/16 who went on to achieve good passes in English and mathematics GCSE
(Timpson Review, 2019, citing DfE matched administrative data for the 2015/16 permanently excluded cohort)
Of adult prisoners who report having been permanently excluded from school.
(Ministry of Justice, Prisoners' Childhood and Family Backgrounds, 2012 — 42% of prisoners reported being permanently excluded from school)
The timing of removals is the signature. They cluster in the run-up to the examination years, which is what one would predict if the driver were the denominator rather than the behaviour. A school whose examination results determine its Ofsted rating and its survival has an institutional incentive to remove pupils whose predicted grades are low before those predictions can become data points. The practice is not confined to the cynical or the unscrupulous; it is the rational response of institutions under existential pressure to the metrics that define their existence.
The apparatus does not merely fail these children. It routes them. Children in alternative provision carry heavy concentrations of special educational needs, social disadvantage and prior adversity (Gill, Quilter-Pinner and Swift, 2017). The pattern of who is routed is not random: excluded children are disproportionately poor, disproportionately in care, disproportionately Black Caribbean boys — the pattern of racial disparity that occasioned the Timpson Review in the first place (Timpson, 2019). The routing is structured along the existing axes of social inequality, which is what one would predict if the driver is constraint load — because social disadvantage raises constraint load independently, so that the children most affected by the institutional environment are also those with the fewest resources to absorb the additional burden.

The strongest quantitative evidence for reading these three curves as one system comes from the literature on exclusion and mental health, which has converged, across independent datasets and methods, on bidirectionality. The bidirectional association is not a nuisance for clean causal inference. Read through the present framework, it is the finding.
Ford and colleagues, analysing the British Child and Adolescent Mental Health Surveys, found that psychological distress predicted later exclusion and that exclusion predicted later distress, including new-onset disorder (Ford et al., 2018). Parker and colleagues found that children with unrecognised and subthreshold difficulties were excluded disproportionately (Parker et al., 2019). In the Avon cohort, children who would go on to be excluded showed elevated difficulties from the age of nine, with further deterioration after exclusion (Tejerina-Arreal et al., 2020).
Fixed-effects analysis of the Millennium Cohort — a design that strips out stable confounding — found the bidirectional association again for emotional symptoms (Villadsen et al., 2024), and inverse-probability-weighted estimates extend the harm into adult health and wellbeing outcomes (Obsuth et al., 2024). These are not weak or preliminary findings. They are consistent across independent datasets, methods and research groups, which in the social sciences represents strong convergence.
British CAMHS surveys: distress predicted exclusion; exclusion predicted new-onset disorder.
Avon cohort: elevated difficulties from age 9 in children later excluded; further deterioration post-exclusion.
Millennium Cohort fixed-effects: bidirectional association for emotional symptoms, controlling for stable confounds.
Inverse-probability-weighted estimates: harm extends into adult health and wellbeing outcomes.
Read atheoretically, bidirectionality is a nuisance — an obstacle to clean causal inference. Read through the present framework, it is the finding. A bidirectional association between distress and removal is what a loop looks like in cohort data: the system detects poor fit, responds by increasing the constraint load on the poorly fitting child, thereby worsens the fit, and reads the worsening as confirmation of the original assessment.
This is Hacking's looping effect operating at institutional speed — the classification changing the classified, the changed behaviour feeding back into the classification (Hacking, 1995; 2007) — with the addition that here the loop has an exit chute. The child who enters the loop as a pupil whose derivation is costly exits it as a permanently excluded young person in alternative provision, carrying a diagnosis that was generated by the institutional environment and will accompany them into every subsequent institutional encounter.
The pupil referral unit is that chute, and the wild-type framework names precisely what it is: institutionalised reference deprivation. Corrigibility — the capacity to test one's take on reality and return to the shared world — is a two-body problem: it requires an available, trusted reference, and insight is a property of the person–reference system, not of the individual alone (Collins, 2026a; 2026b).
Exclusion severs a child's principal daytime relationships with trusted adult reference and with the broad peer reference of a mainstream community. Alternative provision then concentrates reference-deprived children together, in settings where adult reference is scarcer and peer reference is drawn overwhelmingly from others in the same position. The framework's reference-deprivation pathway — capacity to update intact, no trusted external reference available — produces the same phenotype as the neurobiological routes, and this is a mechanistic description of the alternative-provision-to-prison pipeline.
The system takes children whose derivation was costly, removes their reference altogether, and then treats the resulting phenotype as proof of the deficit it began by assuming. The loop closes, and it closes on the children with the least resource to reopen it.
This is not a failure of the system. It is the system operating exactly as its incentive architecture dictates. The PRU is not a safety net. It is the point at which the denominator logic completes itself: the child whose predicted results were a liability has been removed from the data, the school's metrics are protected, and the child's trajectory is set by a configuration of reference deprivation that the sociological literature has documented — without, until now, having a mechanistic account of why it produces the outcomes it does.


The loop operates at institutional speed, through defensible individual decisions, producing a structural certainty: the children who enter it are the children the system was already registering as its most costly members, and the loop reliably makes them more costly — to themselves, to the health system, and eventually to the criminal justice system.
The wild-type hypothesis has always faced a fair evidential demand: interaction claims need interaction evidence. The existing anchors — cross-population genetic studies, foraging environment comparisons — provide suggestive evidence but are methodologically constrained. English education policy since 1988 supplies something these sources cannot: a longitudinal intervention on a single population.
The DRD4 7-repeat allele was associated with better nutritional outcomes among nomadic Ariaal pastoralists and worse outcomes among their recently settled kin — same variant, same people, opposite valence, with environment as moderator. A replication attempt in the neighbouring Rendille did not reproduce the nutritional interaction, which constrains the weight the single finding can bear.
Adults screening positive for ADHD outperformed non-screeners in a patch-foraging environment. The attentional configuration psychiatry treats as deficit confers advantage in the environmental niche it was likely selected in.
The variation itself is argued to be an evolved feature of human adaptability, against the background of the self-domestication literature (Wrangham, 2019; Hare, 2017) and the evolutionary mismatch tradition (Nesse, 2019).
The neurology of English children did not change between 1988 and the present. What changed, in identifiable, dateable steps, was the demand structure and the measurement regime. With each step the reference ecology narrowed further, and with each narrowing the distress, identification and removal curves moved. Same population, same distribution of trait positions, progressively intensified environment, tracked outcomes.
This is the trait-by-environment interaction claim in its strongest available observational form, and it is proof against the type-misreading that has dogged the hypothesis — the mistaken inference that it divides humanity into ancestral and domesticated kinds. In this experiment there are no kinds anywhere in view. Nothing about the children moved. Only the environment did, and the phenotype rates moved with it. The unit of analysis is visibly the pairing, not the individual, not the category.
Same population. Same distribution of trait positions. Progressively intensified environment. Three output curves moving in lockstep. The unit of analysis is visibly the pairing.
The differential susceptibility literature predicts one further feature that the data show: the effects should be unevenly distributed across the sensitivity distribution, concentrated in those most responsive to environmental conditions in both directions (Belsky and Pluess, 2009). The children most damaged by the intensified regime are drawn from the same trait region as the children who flourish most dramatically when conditions are right — which is why the policy conclusion of this analysis is not that expectation or structure harms children, but that a monoculture does.
Vantage sensitivity (Pluess and Belsky, 2013) is the unclaimed dividend of the entire settlement: the capacity for exceptional response to supportive conditions, sitting unexpressed in exactly the children the system is currently routing into its chute. The analysis does not recommend lowered expectation. It recommends ecological diversification — precisely because the children most likely to be damaged by the narrowed monoculture are also the children most likely to generate exceptional outcomes when the conditions are right. They are orchids, not weeds. The current settlement grows no orchids.
It will be objected that the remedy is more rigorous evaluation — better trials of interventions, firmer effect sizes, an education science modelled on evidence-based medicine. The objection misunderstands what the comparative-trial apparatus can see, and psychotherapy research has already run the demonstration.

Half a century of comparative outcome trials has repeatedly failed to separate bona fide therapeutic techniques from one another. The dodo bird verdict — that all have won and all must have prizes — was announced by Rosenzweig in 1936, confirmed meta-analytically by Luborsky and colleagues in 1975, and stands, with qualifications, ever since. Wampold's contextual model draws the conclusion the field keeps resisting: the variance in outcome is carried overwhelmingly by so-called common factors — the alliance, the relationship, expectation, the person of the practitioner — rather than by the technical apparatus the trials were built to compare (Rosenzweig, 1936; Luborsky, Singer and Luborsky, 1975; Wampold, 2015).
In the vocabulary of this paper, the dodo verdict is not an embarrassment but a result: the effective variable in psychological help was always the field conditions — Ground and reflection — and never the technique. The comparative trial cannot detect this because its entire design logic consists of holding the field constant in order to isolate the technique; it is an instrument built to subtract the active ingredient before measuring. The instrument is not defective. It is answering a different question — one whose presuppositions the field has not examined.
The same blindness structures education's "what works" apparatus. An RCT of a phonics scheme or a behaviour programme treats the relational and ecological field of the school as noise to be randomised away, when the accumulated evidence of both fields suggests that the field is the treatment. The noise is the signal. The methodological programme implied here is not anti-empirical; it is that evaluation must take the pairing as its unit — child-in-environment, technique-in-relationship — because evaluating components of a field as if they were drugs is a category error that fifty years of dodo verdicts have already adjudicated.
Most of what circulates as research in this area is not badly executed. It is philosophically unexamined, running a borrowed evidential logic on phenomena whose structure that logic was never built for. The RCT was designed for interventions that work through a mechanism isolable from their relational context. Relationships, fields and ecologies do not have that structure. Applying the RCT to them produces results that are, at best, uninterpretable and, at worst, actively misleading — because a null result for a field-level intervention, measured by a technique-level instrument, is perfectly compatible with the intervention being the most important thing that happened.
Evaluating components of a field as if they were drugs. The pharmacological comparison is doubly confused: it compares two interventions on an entity that may not exist, using an instrument that cannot see what either intervention actually does.
The methodological programme this paper implies: evaluation must take the pairing as its unit — child-in-environment, technique-in-relationship.
Shifting the unit of analysis from the child to the child–environment pairing does not dissolve into utopianism. It generates specific, tractable changes, several of which converge with where the institutional conversation has independently arrived. The framework does not recommend the abolition of structure, assessment or expectation. It recommends a fundamentally different account of what the school is and what it is for.

For assessment, the first question ceases to be what is wrong with this child and becomes: what is this child's position on the acquisition and sensitivity distributions, what is the current demand profile of their environment, and what is the size of the gap? This is an ecological assessment, and its output is a fit formulation rather than a category. The Independent Review's insistence that support follow functional need rather than diagnosis (Fonagy, Wessely and Baird, 2026) is the same move made from the policy side.
The wild-type framework supplies what the Review lacks — an account of why need became transactable only through diagnosis, and therefore what must change for a needs-led system to be stable rather than collapsing back into label economics. The accountability architecture itself must stop making the child a unit of institutional risk, or every reform downstream of it will be metabolised by the same incentive gradient. Changing the assessment instrument without changing the accountability architecture is rearranging the currency while leaving the exchange rate intact.
The restoration of Ground components the measured curriculum crowded out — movement, making, nature, play, unstructured relationship — not as enrichment but as the conditions of capacity expression. These are not optional extras. They are the substrate on which reference builds.
Derivation must become a costed and supported strategy rather than a penalised one. Legitimate modes of acquiring and demonstrating reference must be broadened so that the child who must test, compare and rebuild is not reading their processing cost as evidence of their inadequacy.
Teachers transmitting their own performativity terror downwards (Ball, 2003) are, in field terms, a Ground deficit no child-level intervention can compensate. Adult wellbeing in schools is not a separate policy question. It is the same question, because the field the child inhabits is partly constituted by the adult who inhabits it alongside them.
For the discharge route, the direction is given by the two-body structure of corrigibility: whatever else alternative provision does, it must not deprive a child of trusted reference. This means reintegration designed from day one; mainstream schools retaining accountability for the outcomes of the children they exclude — Timpson's own central recommendation (Timpson, 2019) — and the recognition that concentrating reference-deprived children together is not a neutral placement decision but an active dose of the pathway this paper has described.
The current structure of alternative provision is not an unfortunate side effect of an otherwise functional system. It is the completion of the loop: the accountability apparatus generates the excludable child, the discharge route removes the child from the denominator, and the alternative provision setting administers reference deprivation to a population already reference-deprived. Reforming exclusion policy without reforming the accountability architecture that generates the incentive to exclude is, again, rearranging the downstream while leaving the upstream untouched.
A psychiatry that accepts the referral on the referrer's terms becomes the accountability regime's downstream maintenance department, medicating the canaries so the mine can keep its rating.
The analysis re-describes a large fraction of CAMHS and adult neurodevelopmental demand as referred fit-failure — distress generated in one institution and presented for repair in another. The clinically honest position is to read the distress as signal first, assess the field before the child, use the capacity equation as the differential, and hold the diagnostic act to the standard that it should open the question of what conditions this person needs rather than close it (Collins, 2026a; 2026d).
This is not a counsel of clinical inaction. It is a counsel of clinical honesty. The child in the CAMHS waiting room is not wrong to be distressed. The question is whether the appropriate response is to reduce the child's sensitivity to the environment, or to change the environment — and the answer to that question requires knowing what the environment is and what it is doing, which is precisely the assessment the referral pathway was not designed to perform.
These steps do not replace clinical judgement. They reorient it — from the child as the object of intervention to the child–environment pairing as the unit of analysis, and from the diagnostic category as the destination to the ecological formulation as the starting point.
The argument's force depends on stating plainly what it does not claim. A structural analysis that overstates its claims invites refutation on grounds that are not actually its strongest, and produces policy conclusions that the evidence does not support. The discriminations that follow are not hedges. They are the conditions under which the argument is strongest.
The distribution of trait positions is the explanandum's background, not its product. The argument is about what the environment does to an already-distributed population, not about what produces the distribution.
Distress warrants closer assessment when it is sealed against revision, when sleep, food and relational contact are failing, or when interpretation is accelerating and self-confirming. Signal and symptom require discrimination, not conflation.
Derivation-heavy minds need things to test themselves against. The argument's target is the monoculture and the stakes, not the existence of expectation. Lowered expectation is not the recommendation; ecological breadth is.
The analysis is structural precisely because it explains the pattern without villains. The argument would be weakened, not strengthened, by conspiracy framing. The heads, SENCOs and panels are behaving rationally; the architecture is the problem.
The evidential layers of the argument carry different weights, which should be owned rather than blurred. The mechanism claims — Campbell's law, the incentive analysis of off-rolling, the diagnosis-as-currency account — rest on well-documented institutional behaviour and carry high inferential warrant within their scope. The bidirectional exclusion evidence is quantitatively strong: it is consistent across independent datasets, designs and research groups, and the fixed-effects and inverse-probability-weighted analyses address the most serious confounding concerns.
The attribution of the population distress curve to the accountability regime is ecological and correlational, one plausible major contributor among several, and is asserted at that strength and no more. The conduct-versus-emotional asymmetry in the Millennium Cohort findings (Villadsen et al., 2024) is a live seam: the loop may operate differently for different presentations, and the framework should expect and welcome that differentiation rather than smooth it over. A structural argument that cannot accommodate differential findings has not earned its structural claims.
The framework's own discrimination applies with full specificity. Distress is presumptively signal — the accurate registration of genuine incoherence — when it tracks identifiable conditions in the environment, reduces when those conditions change, and the child retains the capacity for flexible return when conditions are good. These are the markers of a calibrated system responding appropriately to its inputs. The appropriate clinical response is to change the inputs.
Distress warrants closer clinical attention — not as refutation of the signal reading, but as additional hypothesis — when it is sealed against revision regardless of conditions, when it does not reduce when circumstances improve, when sleep, food and relational contact are failing across the board, or when the child's own interpretive frame is accelerating and self-confirming in ways that are not responsive to disconfirmation. These are the markers of a system that may have moved from signal-production into a self-sustaining loop that is no longer tracking the external environment. Even here, the field assessment comes first: some environments are genuinely unrelenting, and persistence of distress under genuinely unrelenting conditions is still signal, not symptom.
The education system built in England after 1988 is the wild-type hypothesis's demonstration case: a single population, an environment intensified in dateable steps, and three output curves — distress, diagnosis, removal — moving together as the reference ecology narrowed.
The same logic that made zero suicide an accountability structure generating coercion and gatekeeping made the league table an accountability structure generating exclusion and label economics. In both cases a phenomenon that emerges from a field was converted into an institutional performance metric, and the institution then reorganised the field to protect the metric. The children whose trait positions paid the highest price for those changes were first recorded as disordered, then transacted as diagnoses, then — when the transaction failed — removed from the denominator and deprived of the very reference relationships through which return remains possible.
Poor fit with the narrowed reference ecology is read as individual deficit. Referral to assessment is generated.
The only currency that moves resources converts fit-failure into a label. The child's account is administratively overwritten.
When the transaction fails, exclusion discharges the liability. The school's metric is protected; the child's trajectory is set.
To call this a mental health crisis in the young is to read the output of the machine as a property of its raw material. The canary data are excellent. It is time to assess the mine.
The rising curves are not evidence of a generation more fragile, more disordered or more in need of clinical management than those before it. They are evidence of an environment that has been moved, in identifiable, dateable steps, further from the conditions that particular trait positions are calibrated for. The canaries are doing exactly what canaries do: they are registering the atmosphere accurately. The response to a canary alarm is not to breed more resilient canaries. It is to check the mine.
The ecological, policy and clinical programme implied by this analysis is tractable. It does not require the abolition of accountability, the abandonment of assessment or the lowering of expectation. It requires that the unit of accountability shift from the measured output of the individual child to the ecological conditions the institution provides for all children; that the assessment apparatus ask the capacity-equation question before it asks the diagnostic question; and that the discharge route be closed by making mainstream schools accountable for the children they exclude. These are not radical demands. Several of them are already in the Timpson Review. What is missing is the structural account of why they keep failing to take hold — and that account is now available.

This paper was developed in dialogue and is jointly authored: the framework, the clinical position, the education analysis and the editorial judgement are Paul Collins's; the drafting, evidence assembly and structural argument were produced by Claude Fable 5 (Anthropic) working within that framework. It belongs to the Spiral Lattice Archive alongside The Untamed Mind, from which the wild-type machinery is drawn.
Spiral Lattice Archive, 27th August 2026.
All works in the Archive share a common framework and are designed to be read in dialogue with one another. Cross-references are substantive, not self-promotional.
Ball, S.J. (2003) 'The teacher's soul and the terrors of performativity', Journal of Education Policy, 18(2), pp. 215–228.
Barack, D.L. et al. (2024) 'Attention deficits linked with proclivity to explore while foraging', Proceedings of the Royal Society B, 291(2024), 20240215.
Belsky, J. and Pluess, M. (2009) 'Beyond diathesis stress: differential susceptibility to environmental influences', Psychological Bulletin, 135(6), pp. 885–908.
Boyce, W.T. and Ellis, B.J. (2005) 'Biological sensitivity to context: an evolutionary–developmental theory of the origins and functions of stress reactivity', Development and Psychopathology, 17(2), pp. 271–301.
Campbell, D.T. (1979) 'Assessing the impact of planned social change', Evaluation and Program Planning, 2(1), pp. 67–90.
Collins, P. (2026a) The Untamed Mind: Wild-Type Cognition and the Domestication Gradient. Spiral Lattice Archive. Available at: https://wild-type-cognition-upeomkz.gamma.site/
Collins, P. (2026b) The Clinical Aperture, Corrigibility, and the Return to the Shared World. Spiral Lattice Archive. Available at: https://clinical-aperture-fjlvu2a.gamma.site/
Collins, P. (2026c) Psychiatry After the Isolated Brain: Mind in a Living World. Spiral Lattice Archive. Available at: https://mind-in-a-living-world-3ntzxiy.gamma.site/
Collins, P. (2026d) Psychiatric Diagnosis as Semiocide. Spiral Lattice Archive. Available at: https://psychiatric-semiocide-vv2kc2s.gamma.site/
Eisenberg, D.T.A. et al. (2008) 'Dopamine receptor genetic polymorphisms and body composition in undernourished pastoralists: an exploration of nutrition indices among nomadic and recently settled Ariaal men of northern Kenya', BMC Evolutionary Biology, 8, 173.
Ellis, B.J. and Boyce, W.T. (2008) 'Biological sensitivity to context', Current Directions in Psychological Science, 17(3), pp. 183–187.
Fonagy, P., Wessely, S. and Baird, G. (2026) Independent Review into Mental Health Conditions, ADHD and Autism: Interim Report. London: HM Government, 31 March 2026.
Ford, T., Parker, C., Salim, J., Goodman, R., Logan, S. and Henley, W. (2018) 'The relationship between exclusion from school and mental health: a secondary analysis of the British Child and Adolescent Mental Health Surveys 2004 and 2007', Psychological Medicine, 48(4), pp. 629–641.
Foucault, M. (1977) Discipline and Punish: The Birth of the Prison. Translated by A. Sheridan. London: Allen Lane.
Gill, K., Quilter-Pinner, H. and Swift, D. (2017) Making the Difference: Breaking the Link Between School Exclusion and Social Exclusion. London: Institute for Public Policy Research.
Hacking, I. (1995) 'The looping effects of human kinds', in Sperber, D., Premack, D. and Premack, A.J. (eds) Causal Cognition. Oxford: Clarendon Press, pp. 351–383.
Hacking, I. (2007) 'Kinds of people: moving targets', Proceedings of the British Academy, 151, pp. 285–318.
Hare, B. (2017) 'Survival of the friendliest: Homo sapiens evolved via selection for prosociality', Annual Review of Psychology, 68, pp. 155–186.
Hutchings, M. (2015) Exam Factories? The Impact of Accountability Measures on Children and Young People. London: National Union of Teachers.
Hutchinson, J. and Crenna-Jennings, W. (2019) Unexplained Pupil Exits from Schools. London: Education Policy Institute.
Luborsky, L., Singer, B. and Luborsky, L. (1975) 'Comparative studies of psychotherapies: is it true that "everyone has won and all must have prizes"?', Archives of General Psychiatry, 32(8), pp. 995–1008.
Ministry of Justice (2012) Prisoners' Childhood and Family Backgrounds. London: Ministry of Justice.
Nesse, R.M. (2019) Good Reasons for Bad Feelings: Insights from the Frontier of Evolutionary Psychiatry. New York: Dutton.
Obsuth, I. et al. (2024) 'The impact of school exclusion in childhood on health and well-being outcomes in adulthood: estimating causal effects using inverse probability of treatment weighting', British Journal of Educational Psychology, 94(3).
Parker, C., Tejerina-Arreal, M., Henley, W., Goodman, R., Logan, S. and Ford, T. (2019) 'Are children with unrecognised psychiatric disorders being excluded from school?', Psychological Medicine, 49(15), pp. 2561–2572.
Pellicano, E. and Burr, D. (2012) 'When the world becomes "too real": a Bayesian explanation of autistic perception', Trends in Cognitive Sciences, 16(10), pp. 504–510.
Pluess, M. and Belsky, J. (2013) 'Vantage sensitivity: individual differences in response to positive experiences', Psychological Bulletin, 139(4), pp. 901–916.
Rosenzweig, S. (1936) 'Some implicit common factors in diverse methods of psychotherapy', American Journal of Orthopsychiatry, 6(3), pp. 412–415.
Strathern, M. (1997) '"Improving ratings": audit in the British university system', European Review, 5(3), pp. 305–321.
Taylor, H., Fernandes, B. and Wraight, S. (2022) 'The evolution of complementary cognition: humans cooperatively adapt and evolve through a system of collective cognitive search', Cambridge Archaeological Journal, 32(1), pp. 61–77.
Tejerina-Arreal, M. et al. (2020) 'Child and adolescent mental health trajectories in relation to exclusion from school from the Avon Longitudinal Study of Parents and Children', Child and Adolescent Mental Health, 25(4), pp. 217–223.
Timpson, E. (2019) Timpson Review of School Exclusion. CP 92. London: Department for Education.
Van de Cruys, S. et al. (2014) 'Precise minds in uncertain worlds: predictive coding in autism', Psychological Review, 121(4), pp. 649–675.
Villadsen, A. et al. (2024) 'Longitudinal association of conduct and emotional problems with school exclusion and truancy: a fixed effect analysis of the UK Millennium Cohort Study', Child and Adolescent Mental Health, 29(2).
von Uexküll, J. (1934/2010) A Foray into the Worlds of Animals and Humans. Translated by J.D. O'Neil. Minneapolis: University of Minnesota Press.
Wampold, B.E. (2015) 'How important are the common factors in psychotherapy? An update', World Psychiatry, 14(3), pp. 270–277.
Wrangham, R. (2019) The Goodness Paradox: How Evolution Made Us Both More and Less Violent. London: Profile Books.