Summary of Findings

The 119 entries in the False Assumption Registry share one structural feature above all others: the error was not random. In case after case, the false belief was the one that served the interests of the institutions promoting it. Pharmaceutical companies and psychiatrists promoted the serotonin hypothesis of depression because it sold drugs and simplified clinical conversations. Diversity consultants defended anti-bias training because their industry depended on it. Intelligence officials promoted WMD certainty because the policy was already decided. Education schools taught three-cueing because phonics instruction required admitting that decades of teacher training had been wrong. The pattern is not that experts lied, exactly, but that the incentive structure reliably pointed toward one conclusion, and experts found reasons to believe it. What looks like an epistemological failure is, underneath, an institutional one: no mechanism existed to make the cost of being wrong fall on the people who were wrong. The costs landed instead on autistic children denied effective reading instruction, on elderly people deprioritized in vaccine queues, on girls in Rotherham, on soldiers in Iraq.

A second pattern runs through nearly every entry: moral framing was used as a substitute for empirical engagement. Questioning the sentencing disparity literature was treated as defending racism. Questioning facilitated communication was treated as denying agency to disabled people. Questioning gender-affirming care protocols was treated as wanting children to suffer. Questioning the "mass graves" framing at Canadian residential schools was treated as minimizing colonial violence. In each case, the moral stakes of the question were elevated until empirical challenge became reputationally dangerous, and the question stopped being asked. This is distinct from ordinary motivated reasoning. It is a deliberate rhetorical move, sometimes conscious and sometimes not, that converts a factual dispute into a moral one, where the person raising evidence is the problem. The registry documents the consequences: the moral shield held long enough for the harm to accumulate. The Rotherham abuse continued for sixteen years. The amyloid hypothesis consumed thirty years of Alzheimer's funding. Children failed to learn to read for a generation. The duration of harm tracks almost perfectly with how effectively the moral-framing defense insulated the assumption from challenge.

A third pattern concerns the manufacture and weaponization of consensus. Across dozens of entries, a small number of influential papers or figures established a position that was then cited as self-reinforcing authority. Two papers effectively closed the COVID lab-leak debate for two years. Ancel Keys selected seven countries from twenty-two to build the saturated fat hypothesis, then shaped the American Heart Association's guidelines, which shaped federal dietary policy, which shaped what doctors told patients for fifty years. Each institutional endorsement was treated as independent confirmation when it was actually serial citation of the same original claim. The Lancet letter on COVID origins, the WPATH guidelines on gender care, the Fajnzylber papers on inequality and crime, the Dunn and Dunn learning styles inventories: all became load-bearing structures that entire professional ecosystems were built around, which meant that questioning the foundation threatened the building. Peer review, which exists precisely to catch this kind of error, failed consistently, partly because reviewers were drawn from the same intellectual communities that produced the original claims, and partly because journals rewarded novelty and confirmation over replication and challenge.

What the registry documents, taken whole, is the near-total absence of internal correction. In 119 entries, there is not a single clear case of an institution proactively identifying its own false assumption and correcting it before external pressure forced the issue. Correction came from investigative journalists, from foreign health agencies (Sweden, Finland, and the UK's Cass Review reversed course on pediatric gender medicine while American professional bodies held firm), from election results that falsified strategic predictions, from whistleblowers, from replication studies published in journals that had ignored the original methodological objections for years. The correcting voice was almost always available before the harm peaked; the institutional failure was not lack of information but the systematic underweighting of dissenting information. This is the registry's central finding, and it is not a finding about bad people. It is a finding about structures: about who bears the cost of error, about what moral framing can do to empirical inquiry, about how prestige substitutes for evidence, and about the remarkable capacity of institutions to protect themselves from the knowledge that they are wrong.

The most consistent pattern across all 119 entries is not ideological: it is structural. Institutions adopt a belief, build programs around it, and then defend those programs long after the belief has been falsified, because the programs have become the institution. The dietary guidelines kept Americans eating refined carbohydrates for decades after the saturated fat hypothesis collapsed, because the Food Guide Pyramid was already printed and distributed and the agencies that produced it had no mechanism for admitting error. The Luevano Consent Decree banned valid federal hiring exams in 1981 and left agencies selecting employees on inferior instruments for forty-four years, because no one in the system had both the incentive and the authority to reopen the question. The amyloid hypothesis consumed roughly half of Alzheimer's research funding for a generation, because the FDA had embedded amyloid clearance as an accepted surrogate endpoint, making the hypothesis legally and bureaucratically load-bearing. In each case, the false assumption did not merely produce a bad policy; it produced an institution that required the assumption to remain true. Correction, when it came, required either a catastrophic visible failure or an external political shock, neither of which is a reliable mechanism for governing a complex society.

Several of the most damaging present-day crises trace directly through this pattern. The homicide surge that began in 2020 is downstream from at least three compounding false assumptions: that poverty, not enforcement, drives urban violence; that policing disparities prove discrimination rather than reflecting offense patterns; and that reducing proactive enforcement would not increase victimization. Each assumption produced a policy layer, and the layers accumulated. Stop-and-frisk ended in 2013. Prosecutorial discretion narrowed. Consent decrees restructured department protocols. When the Floyd-era de-policing accelerated all of these trends simultaneously, the result was an estimated five thousand additional murders in 2020 alone, with Black Americans bearing roughly half the toll. The political response has been to debate the framing of the 2020 spike rather than the thirty-year policy architecture beneath it, because acknowledging that architecture would require admitting that the assumptions driving it were wrong, and those assumptions are still institutionally defended. Baltimore's clearance rate sits below fifty percent, with over fifteen hundred unsolved killings. The causal chain from false assumption to dead Black residents is not abstract; it is documented and specific, and the policy apparatus that produced it remains largely intact.

A second major pattern is what might be called the equity override: the systematic choice to prioritize distributional optics over evidence, with the result that the populations the policies claimed to protect bore the costs. The CDC's vaccine prioritization in late 2020 moved essential workers ahead of high-mortality seniors, and the agency's own modeling projected a seven percent increase in deaths from the chosen plan. Test-blind admissions, sold as a tool for disadvantaged students, made high-ability low-income students less likely to be identified, because inflated GPAs replaced the one signal that cut through socioeconomic noise. The Luevano Decree banned written civil service exams on adverse-impact grounds and replaced them with subjective resume reviews that introduced different biases while degrading overall hiring quality. Foreign aid to sub-Saharan Africa, disbursed at a scale equivalent to four Marshall Plans over four decades, coincided with declining per-capita GDP and accelerating capital flight. The grooming gang scandal in Britain is perhaps the starkest case: the assumption that discussing perpetrator ethnicity constituted a moral panic led police and social services to suppress evidence for sixteen years, producing an estimated ten thousand or more victims, the overwhelming majority of them working-class girls from the communities that progressive policy claimed to protect. In each instance, the equity framing was not incidental to the harm; it was the mechanism that delayed correction, because questioning the policy could be characterized as opposing the goal.

What the registry reveals about how societies correct course is not encouraging. Correction tends to arrive in one of two forms: slow accumulation of undeniable failure, or political backlash that overshoots. The replication crisis eventually undermined the implicit association test, stereotype threat research, and the ego depletion literature, but the mandatory training programs built on those findings are still running, and the political reaction to them is now sweeping away potentially legitimate civil rights infrastructure alongside the pseudoscientific additions. The 2025 federal DEI rollback does not distinguish between programs that worked and programs that did not, because the backlash is not an evidence-based correction; it is a political one. The same dynamic is visible in gender medicine, where the overcorrection in twenty states banning all interventions for minors followed years of overcorrection in the opposite direction, with neither phase adequately grounded in clinical evidence. The suppression entries are particularly instructive here: the longer an inconvenient finding is suppressed, the more extreme the eventual reckoning. Sixteen years of grooming gang suppression produced a political crisis that a prompt institutional response in 2005 might have contained. Five years of COVID origin suppression produced durable damage to public trust in scientific institutions that honest early inquiry might have avoided. The registry does not document a society that learns from false assumptions. It documents a society that accumulates them until the weight becomes politically unsustainable, and then lurches.

The most consistent finding across all 119 entries is that federal agencies and elite academic institutions are the primary engines of false assumption propagation, and they rarely operate independently. The pattern is nearly universal: a finding originates in a narrow academic study, passes through a prestige journal, receives amplification from a federal agency or professional body, and then becomes embedded in policy before any serious replication effort occurs. The NIH funded amyloid research for decades while career structures punished dissent. The CDC's ACIP embedded equity-over-mortality logic into vaccine allocation during a supply-limited emergency. The EEOC institutionalized test-skepticism through litigation and consent decrees that lasted 44 years. The FBI's Uniform Crime Reporting program published imputed figures as authoritative national data while the Bureau of Justice Statistics quietly published critiques of the same methodology. What distinguishes government agencies from other institutional actors is their capacity to make assumptions compulsory: through funding conditions, regulatory mandates, consent decrees, and congressional directives, they convert contested empirical claims into operational requirements that practitioners must follow regardless of the underlying evidence.

Professional associations and prestige journals perform a distinct but equally important function: they manufacture the appearance of consensus. The American Medical Association, the American Psychological Association, the American Anthropological Association, the World Professional Association for Transgender Health, and the Association of American Medical Colleges all issued formal statements or policy positions that treated contested empirical questions as settled, then used membership norms, accreditation requirements, and licensing structures to enforce compliance. Journals including Nature, PNAS, the Lancet, Psychological Review, and the American Journal of Psychiatry published methodologically weak findings that gained authority through venue association alone. The peer review process, far from serving as a corrective mechanism, repeatedly functioned as a credibility-laundering operation: low inter-rater reliability, reviewer selection from familiar professional networks, and publication bias toward positive findings meant that the imprimatur of peer review added prestige without adding scrutiny. When dissenters did publish, the response was frequently professional rather than scientific: Norman Wang's critique was retracted after colleagues called for his dismissal; Lilienfeld's microaggression paper drew accusations of minimizing racism rather than methodological engagement; Amy Wax faced formal university proceedings classified as unprofessional conduct rather than academic dispute.

Responsibility is diffuse rather than concentrated, but certain structural positions recur with enough regularity to constitute a pattern. The charismatic individual with an institutional platform appears across entries spanning decades and disciplines: a single clinician, researcher, or advocate converts personal authority into institutional infrastructure, which then attracts government funding, which then makes reversal politically costly. Van der Kolk built a trauma center, wrote a bestseller, and ended up with a Congressional mandate for 150 federally funded clinical centers. Kübler-Ross's clinical observations about dying patients became a global standard for grief through a bestseller and a Time 100 listing, with no replication required. Biklen imported Facilitated Communication from Australia to Syracuse University and created a national training program before controlled experiments showed facilitators were authoring the output themselves. In each case, the individual's credibility fused with institutional infrastructure to produce an assumption that outlasted any subsequent disconfirmation. A related pattern involves individuals whose prior credible work lent legitimacy to later claims that did not deserve it: Steele's FIFA corruption work made his dossier credible to the FBI; Cheney's Gulf War experience made his WMD assertions credible to Congress, even as his own 1994 public statements directly contradicted them.

Media organizations, think tanks, and advocacy groups complete the circuit. The New York Times appears as an enabling institution in a remarkable share of entries, not through random error but through systematic editorial choices: suppressing crime-context data while reporting disparity statistics, correcting factual errors quietly without acknowledgment, and framing empirical disagreement as moral failure. The Associated Press enforced social-construct language through style guide directives that propagated across thousands of newsrooms. Think tanks including the American Enterprise Institute, the Council on Foreign Relations, and the Project for the New American Century converted contested empirical claims into actionable policy positions faster than peer review could operate, then fed those positions back to prestige media as expert consensus. Advocacy organizations including the ACLU, the Good Law Project, and the NAACP Legal Defense Fund embedded contested empirical claims in constitutional litigation, where they became harder to dislodge through normal evidentiary processes. The same names and organizations do not always recur across entries, but the same institutional types do, operating through the same mechanisms: funding as enforcement, prestige as credibility laundering, moral framing as scrutiny suppressor, and professional punishment as the last line of defense when evidence accumulates against the assumption.

The most consistent difference between those who believed the false assumptions and those who suspected them is not intelligence, domain expertise, or access to data. It is a prior question: what counts as sufficient grounds for belief? Believers, across domains as different as nutritional science, criminal justice, developmental psychology, and foreign policy, treated narrative plausibility as a stopping point. If a story was internally coherent, morally satisfying, and endorsed by a prestigious institution, that was enough. The refrigerator mother theory required no neurological mechanism because the psychoanalytic story was complete on its own terms. The chemical imbalance theory of depression required no direct measurement of serotonin deficits because drug efficacy made the deficit narrative feel confirmed. The 1619 Project required no engagement with competing historical evidence because the Pulitzer Prize had already validated it. Skeptics, by contrast, treated narrative coherence as the beginning of inquiry, not its end. They asked what the specific causal mechanism was, whether the measurement instrument had been validated against an independent criterion, and whether the evidence would survive a preregistered replication. This is not a difference in cleverness. It is a difference in what the mind treats as a closed question.

A second structural difference concerns the relationship between moral stakes and evidentiary standards. Believers consistently lowered the bar for claims that confirmed their moral framework and raised it for claims that threatened it. This pattern appears across every domain in the registry. The mass graves at Canadian residential schools were reported as confirmed before any excavation, because the history of residential schools was genuinely terrible and the specific claim felt like a natural extension of that truth. Stereotype threat research accumulated six hundred supportive studies in print while null results stayed in file drawers, because the theory aligned with a preference for situational over biological explanations of achievement gaps, and skepticism could be classified as indifference to racial injustice. The Lancet letter calling the lab-leak hypothesis a conspiracy theory was signed by scientists with undisclosed conflicts of interest and treated as dispositive, because the alternative carried diplomatic and institutional costs that no one wanted to bear. Skeptics did not lack moral commitments. What they refused to do was allow moral urgency to set the evidentiary threshold. Marc Lipsitch modeled vaccine prioritization outcomes within every racial and ethnic group and found that age-first allocation saved more lives including within minority populations. The finding was inconvenient for the equity framing, and he published it anyway. That willingness to follow the data to an uncomfortable conclusion is the defining characteristic of the skeptic across the registry.

The believers also share a specific relationship to institutional authority that the skeptics do not. Believers used prestige as a proxy for validity. A Nobel Prize made lobotomy standard practice. A Harvard affiliation and Pulitzer-winning prose gave Stephen Jay Gould's reanalysis of Morton's skulls thirty years of unexamined life. The Endocrine Society's endorsement of pediatric gender transition was cited in court briefs as settled science. The FBI's Uniform Crime Report carried the implicit authority of a federal statistical agency even when its underlying data was imputed rather than observed. In each case, the credential or the institutional imprimatur substituted for examination of the underlying methodology. Skeptics consistently went one level deeper. Lewis and colleagues actually remeasured Morton's skulls rather than accepting Gould's reanalysis of published tables. Borjas and VerBruggen reanalyzed the identical Florida dataset on physician race and newborn outcomes and found the omitted variable that the original authors had missed. DARPA evaluated the specific proposed experiment in the DEFUSE protocol, the furin cleavage site insertion at a BSL-2 facility, rather than accepting the NIH's categorical judgment that gain-of-function research was comparable to other biomedical advances. The skeptic's relationship to authority is not contrarian. It is conditional: authority is evidence, not proof.

The believers also share a particular model of human nature and social causation that the skeptics tend not to share. Across the registry, believers gravitated toward blank-slate environmentalism, structural determinism, and teleological history. If outcomes differed between groups, the cause was environmental, and the environment was controllable. If institutions were reformed, outcomes would equalize. If history had a direction, the present moral consensus would be vindicated by future judgment. These commitments were not merely political. They were load-bearing assumptions that made certain empirical questions feel already answered. The Coleman Report's 1966 finding that family background dominated school resources was largely ignored for sixty years not because educators failed to read it, but because it contradicted a premise that the entire policy apparatus depended on. The behavioral genetics literature showing roughly eighty percent heritability for adult IQ in wealthy countries was not engaged; it was suppressed, because the blank-slate model was not a hypothesis to be tested but a precondition for a certain kind of politics. Skeptics were not uniformly on the other side of these questions. What distinguished them was that they treated these as open empirical questions rather than settled moral ones. Bernard Rimland proposed neurological causation for autism in 1964 not because he was indifferent to suffering but because he had a child with autism and the psychogenic theory had no mechanism he could identify. The biological hypothesis was not ideologically comfortable. It was simply more consistent with the evidence he could examine.

Finally, the registry reveals a consistent pattern in who gets fooled versus who does not, and it is not primarily a matter of credentials. The believers were disproportionately embedded in the institutional structures that produced and reproduced the false assumptions: the psychoanalytic training system, the NIH grant network, the teacher education apparatus, the corporate DEI consulting industry, the academic journals with publication bias toward positive findings. The skeptics were disproportionately outsiders, practitioners, or people with adversarial institutional positions. Rimland was a parent, not a psychiatrist. DARPA was outside the NIH funding structure. The economists who caught the birth-weight omission in the physician-race study were not medical researchers. The cognitive scientists who falsified the three-cueing reading system were not embedded in teacher education. This is not because outsiders are smarter. It is because insiders face a specific epistemological hazard: their funding, their credentials, their professional relationships, and their sense of moral identity are all tied to the assumption being true. The suppression-perpetuation loop that runs through the registry, where concealing disconfirming evidence extends the lifespan of false assumptions, is not primarily a story about dishonesty. It is a story about what happens when the cost of being wrong is borne entirely by people outside the institution, while the cost of admitting error is borne entirely by people inside it.

The 119 entries in the False Assumption Registry share a structural problem that precedes any specific methodological failure: the institutions producing the evidence were also the primary beneficiaries of the conclusions. NIAID administered grants to researchers whose work justified gain-of-function experiments while Fauci simultaneously defended the practice in print. The Information Technology Association of America generated STEM shortage data while lobbying for the H-1B expansions that shortage data justified. EcoHealth Alliance's Peter Daszak organized the Lancet letter dismissing the lab-leak hypothesis while his organization had direct financial ties to the Wuhan Institute of Virology. McKinsey and BCG produced the studies showing diversity correlates with financial performance while selling diversity consulting to the same corporations. The pattern is not that interested parties occasionally influence research; it is that in these cases, the interested party was the research infrastructure. A practical heuristic follows directly: before accepting any empirical claim, identify who funded the primary studies, who conducted them, and who benefits institutionally from the conclusion. When those three answers point to the same entity, independent replication is not a nicety but a prerequisite.

The evidentiary foundations of these assumptions were almost uniformly thinner than their public authority suggested. Kanner's 1943 paper describing eleven children became the clinical basis for blaming mothers for autism across three decades. The John/Joan case, a single subject whose outcome was actively misreported, became the experimental proof of gender as a social construction. The 2006 Nature oligomer paper, later found to contain manipulated images, anchored roughly half of NIH's Alzheimer's budget for fifteen years. Fewer than ninety Stanford children, drawn from an unusually stable and educated population, became the marshmallow test's claim on self-control research worldwide. The specific warning sign here is not merely small samples but the citation pattern that follows: when subsequent literature cites the original study rather than independent replications of it, and when the citation count grows while the replication count stays near zero, the field has built a pyramid on a single brick. Ego depletion accumulated over six hundred supportive studies before a large preregistered replication found nothing; the volume of positive findings was treated as confirmation rather than as a reason to ask why no one had run the definitive test. A claim supported by many studies that all trace back to one research group, one methodology, or one dataset is not a claim supported by many studies. It is one claim, cited repeatedly.

The propagation of these assumptions followed a consistent institutional chain that a careful observer could have traced in real time. An academic paper with a methodological flaw appears in a prestigious journal. Science journalists report the finding without engaging the methods section. Advocacy organizations incorporate it into policy briefs. Government bodies or courts cite the advocacy briefs as authoritative. By the time the correction arrives, the assumption is embedded in clinical guidelines, legal precedent, school curricula, or corporate compliance infrastructure, and the correction must fight not just the original error but the entire institutional apparatus that has grown up around it. The PNAS study on Black doctors and Black newborns traveled from journal to Nature commentary to AAMC amicus brief to Justice Jackson's Supreme Court dissent; the AAMC sent a clarification letter after oral arguments, too late to affect the judicial record. The 1619 Project moved from New York Times Magazine to Pulitzer Prize to K-12 curricula; the project's website quietly removed the phrase "our true founding" in September 2020 with no correction notice. The speed asymmetry is the key diagnostic: when the infrastructure for spreading a claim (prestige media, policy networks, professional associations) is vastly more efficient than the infrastructure for correcting it (peer review, archival research, FOIA requests), the claim will be institutionally embedded long before its weaknesses are publicly visible.

Across the registry, critics raised methodological objections early, often within years of the original claim, and were answered with social sanction rather than empirical counter-argument. Rimland published a systematic critique of the refrigerator mother hypothesis in 1964; clinical practice did not change for another decade. Arrow and Capron questioned the STEM shortage narrative as early as 1959; Congress expanded H-1B repeatedly anyway. Ebright and Chan warned about gain-of-function risks before COVID; their warnings received no institutional traction. The Canadian Archaeological Association stated from the beginning that ground-penetrating radar identifies soil disturbances, not bodies, and that excavation was required before conclusions could be drawn; the warning was ignored while "215 children" circulated globally as a confirmed figure. The specific pattern to watch for is not the existence of dissent but the form of the response to it. When critics are answered with data, the field is functioning. When they are answered with labels, the field is protecting itself. Lab-leak proponents were called conspiracy theorists in a Lancet letter. Researchers who noticed racial patterns in crime statistics were labeled fringe extremists. Scholars who questioned transgender youth treatment protocols were accused of contributing to genocide. Economists who questioned the STEM shortage were ignored while their work was available in standard databases. In each case, the label did the work of rebuttal, and the absence of an empirical counter-argument was itself the signal that no adequate counter-argument existed.

Several of these assumptions survived as long as they did because questioning the empirical claim felt morally equivalent to endorsing a harm. After Brown v. Board, challenging the Clark doll test's methodology felt like challenging the moral legitimacy of desegregation itself. Mentioning the ethnicity of grooming gang perpetrators was equated with helping the British National Party. The AAMC required unconscious bias training at all 155 medical schools within days of George Floyd's death, before the evidence base for such training had been established, because the speed of adoption was driven by moral urgency rather than empirical readiness. The IAT's test-retest reliability was known to be low before it was institutionalized; Forscher et al. showed in 2019 that changing IAT scores did not change behavior; none of this slowed adoption because the training had become a moral signal rather than an intervention. The practical test is this: when you ask what evidence would falsify a claim and the answer is that no evidence could, because doubting the claim is itself evidence of bad character, the claim has left the domain of empirical inquiry. Proponents of the "wrong side of history" framework never specified what would count as evidence that history had no discernible direction. Advocates of the chemical imbalance theory of depression never specified what serotonin level would constitute a deficit. The unfalsifiability was not a subtle philosophical problem; it was visible in the structure of the argument.

The registry also reveals a consistent pattern of selective data use that a careful reader of the original studies could have detected. The sheepskin effect, showing that graduation year produces a 34 percent earnings premium against 5 percent per normal year, was present in the same datasets used to support human capital theory but was not integrated into the dominant model. Wage stagnation for engineers and software developers was available in standard labor statistics throughout the period of STEM shortage claims; basic economics predicts rising wages in a genuine shortage, and the wages were not rising. The Seven Countries Study selected countries that fit the fat-heart disease hypothesis; data from countries contradicting it were available but not incorporated, a fact that Canadian and British health authorities noted almost immediately after the 1977 Senate report. The New York Times published a graph plotting 2020 homicides as if they occurred on January 1, converting a temporal question about when the surge began into a visual artifact; CDC weekly mortality data showing the surge began the week of George Floyd's death was publicly available. In each case, the selective presentation was not subtle. The omitted variable, the missing control, the aggregated statistic that disaggregated differently, the per-capita figure that contradicted the total, were present in adjacent literature or the same dataset. The question to ask of any empirical claim is not only what data supports it but what data the proponents are not discussing, and whether the omission is explained or merely convenient.

A final pattern runs through the registry that functions as a summary heuristic. When a new consensus forms rapidly, when institutional adoption precedes replication, when critics are answered with labels rather than data, when the primary producers of evidence are also the primary beneficiaries of the conclusion, when no one can articulate what evidence would falsify the claim, and when the real-world outcome test is being failed and explained away rather than treated as disconfirming, the assumption is almost certainly being held in place by something other than evidence. The something is usually a combination of funding streams, career incentives, moral framing, and the simple social cost of being the person in the room who asks the obvious question. The obvious question, asked early enough, would have been sufficient in most of these cases. Why has no one replicated this? Who funded the study? What does the per-capita figure show? What would falsify this? Why are the critics being fired rather than answered? These are not sophisticated methodological questions. They are the questions a careful newspaper reader could ask. The registry's most consistent finding is that they were not asked, not because they were difficult, but because asking them was costly.

The causal chain the thesis proposes — expert claims harden into common knowledge, common knowledge drives policy, policy causes harm — holds up across a substantial majority of the 119 entries. The fit is strong or at least partial in roughly 85 to 90 of them. The clearest cases are also the most consequential: the vaccine allocation debate over essential workers versus seniors, where the CDC's own modeling showed the equity-first approach would increase deaths by up to seven percent; the grooming gangs suppression, where the Jay Report documented over 1,400 victims in Rotherham alone; the lobotomy epidemic, where Freeman's road tours normalized 40,000 to 50,000 procedures before any controlled trial was run; and the reading wars, where Goodman's "psycholinguistic guessing game" theory became classroom orthodoxy and produced measurably the worst adjusted reading scores in Oregon's history. In each case the chain is traceable, the links are documented, and the harms are concrete. Where the chain breaks down, it usually breaks at the policy link: entries like Oliver Sacks' storytelling, the Race-IQ taboo, and the income-inequality-drives-crime claim produce epistemic and institutional harms rather than identifiable policy failures with countable victims. A handful of entries — Black on White Crime, Race as Entirely a Social Construct, the 1619 Project — have harm claims that are either anecdotal, contradictory, or themselves contested on the same empirical grounds the entry is supposed to adjudicate. These are the weakest fits, and they are worth noting honestly.

The deeper pattern — moral conviction disguised as empirical claim — is visible in a clear majority of entries, perhaps 65 to 75 of the 119. The pattern is most explicit where the moral commitment preceded the research program and shaped which evidence was gathered, which anomalies were ignored, and what counted as a legitimate objection. The vaccine allocation entry is the starkest: the committee's own mortality modeling contradicted the equity argument, and the equity argument won anyway. The grooming gangs entry shows the same structure: anti-racism conviction caused authorities to relabel victims as "child prostitutes" and suppress ethnic pattern data. The facilitated communication entry is a textbook case of a moral conviction — disabled people have hidden normal intelligence and deserve to be heard — generating outright empirical fraud, with facilitators unconsciously steering responses and researchers resisting falsification because the conclusion felt morally necessary. Lysenko and Prohibition are the historical archetypes: Lamarckism was politically convenient for Marxism's rejection of bourgeois randomness, and temperance was a Protestant moral crusade stated as a public health causal claim. The trans-skepticism-causes-suicide entry is perhaps the purest modern instance: "skepticism equals genocide" has zero empirical content; it is a moral claim that was deployed as a public health finding to suppress clinical dissent. The implicit bias test, anti-bias training, stereotype threat, and microaggressions entries all share the same structure: a moral conviction about pervasive unconscious racism preceded and shaped research programs that then failed to replicate, but whose policy applications were already institutionalized before the replication failures were acknowledged.

That said, a meaningful minority of entries — perhaps 25 to 35 — are better explained as straightforward empirical errors, methodological failures, or industry capture, with moral conviction playing a secondary or negligible role. The housing bubble entry is the clearest: Mayer, Sinai, and Smith used regression models and rent-price ratios; the error was methodological, not normative. The peer review entry is similar: the belief in inter-rater reliability was a quality-control assumption, not a values claim. Myers-Briggs is a psychometric validity failure driven by commercial incentives. The amyloid cascade hypothesis is a genuine scientific error compounded by a fraudulent 2006 paper and funding inertia; the moral element is essentially absent. The ego depletion, marshmallow test, grit, and learning styles entries cluster together as products of the replication crisis — small samples, publication bias, p-hacking — rather than moral conviction. The STEM shortage claim is primarily industry lobbying dressed as workforce analysis. These entries fit the first part of the thesis (the causal chain) but not the second (the moral disguise). They suggest that the thesis, while capturing a real and important pattern, overstates its universality if presented as the dominant explanation for all false assumptions in the registry.

The thesis also misses several recurring patterns that deserve independent explanation. Financial incentives are underweighted. The chemical imbalance theory of depression was amplified by pharmaceutical advertising, not primarily by moral conviction about the nature of suffering. Media consolidation was driven by regulatory capture and economic ideology, not by a values commitment to broadcasting quality. The Blue Zones longevity claims were sustained by a publishing and wellness industry with direct commercial stakes in the narrative. Guild protectionism explains the specialist-therapy entry better than moral conviction does: licensing structures protected professional territory, and the SUMMIT trial's demonstration that nonspecialists suffice was resisted for reasons of economic self-interest. Bureaucratic inertia is also underweighted: the amyloid hypothesis persisted partly because NIH funding structures reward incremental work within established paradigms, not because researchers had a moral commitment to amyloid. And genuine scientific uncertainty, as distinct from motivated reasoning, explains some entries better than the thesis allows: the immigration-compensates-for-low-birth-rate claim involves contested economics where reasonable experts disagree; the universal health standards entry reflects overgeneralization from European samples, a methodological error without a clear moral driver. The thesis also does not account for cases where the false assumption originated in popular belief rather than expert claims: the five stages of grief became clinical doctrine partly because it resonated with lay experience before it was formally adopted; the "wrong side of history" trope was political rhetoric that academics later rationalized rather than an expert claim that filtered down.

One pattern the thesis identifies correctly but understates is the enforcement mechanism. Across a large number of entries, what distinguishes a moral conviction from a merely mistaken empirical claim is not the content of the belief but the response to challenge. James Watson's career was destroyed for raising the IQ gap question. Researchers studying group differences in athletic ability face institutional retaliation. The trans-skepticism entries document career penalties for clinicians who raised questions about pediatric gender medicine. The Race-IQ taboo entry notes that the taboo is enforced by moral stigma, not by evidence. This enforcement pattern is diagnostic: genuine empirical disputes are settled by data; moral boundary violations are settled by exclusion. The thesis is right that this is a distinctive failure mode, and the registry provides substantial evidence for it. What the thesis does not fully reckon with is that some entries in the registry appear to be subject to the same failure mode from the opposite direction: the Black on White Crime entry, the Race as Social Construct entry, and arguably the Skull Measurements entry (which treats Gould's anti-racist reanalysis as the false assumption) show signs of moral conviction driving empirical claims in a direction the registry's compilers apparently find more congenial. This is worth acknowledging, because a theory about moral convictions distorting empirical claims should be applied consistently.

The overall verdict is that the thesis is substantially supported by the data, with important qualifications. The causal chain holds for a large majority of entries and is well-documented in the most consequential cases. The moral-conviction-as-empirical-claim pattern is real, widespread, and explains something important about why these false assumptions were resistant to correction: they were protected not by evidence but by the social cost of questioning them. The thesis is weakest in two respects. First, it overgeneralizes: a significant minority of entries are better explained by methodological failure, financial incentives, or bureaucratic inertia, and the thesis needs to account for these without forcing them into the moral-conviction template. Second, the thesis's historical explanation — that Western societies lost teleological and natural law frameworks after WWII, leading moral commitments to be wired into empirical claims — is plausible for the post-1945 entries but is not tested by the data in any rigorous way. Lysenko, phrenology, Prohibition, and primitive communism all predate or are contemporaneous with the WWII watershed and show the same pattern, which suggests the failure mode is older and more general than the thesis's historical account implies. The pattern may be less a product of a specific post-war philosophical rupture and more a standing feature of how moral communities use empirical language to insulate normative commitments from scrutiny. That is a less dramatic thesis, but it fits the full range of evidence better.