On 28 February 1998, gastroenterologist Andrew Wakefield stood before cameras at London’s Royal Free Hospital and told the public that the measles-mumps-rubella (MMR) vaccine should be given as three separate jabs, citing a Lancet paper he had co-authored on twelve children. The paper itself never claimed to prove that MMR caused autism — its discussion section explicitly conceded “We did not prove an association between measles, mumps, and rubella vaccine and the syndrome described.” The promise of the press conference and the reality of the data were therefore divorced from the first day: a hypothesis dressed as a finding, a twelve-patient case series dressed as a public-health alarm. The gap would be measured, over the following two decades, in collapsed vaccination rates, returned outbreaks of an eliminated disease, and at least one young man dead of measles pneumonia in a country that had not recorded such a death in years.
The paper, titled “Ileal-lymphoid-nodular hyperplasia, non-specific colitis, and pervasive developmental disorder in children,” survived twelve years as a live citation before it was fully retracted by The Lancet on 2 February 2010, and its lead author was erased from the UK medical register on 24 May 2010. What the intervening investigation — driven not by a regulator but by Sunday Times journalist Brian Deer — established was not honest error but fabrication. Of the twelve children’s case histories, not one matched the medical records cleanly; histopathology reported in the paper as “non-specific colitis” had in several cases been recorded as normal or unremarkable by the Royal Free’s own pathologists and altered upstream toward a diagnosis the thesis required. Children whose regression the paper timed to the days after vaccination had records placing the onset before the jab, or months after it.
The motive was financial and concealed. Wakefield had been paid £435,643, plus a £55,000 grant, via solicitor Richard Barr and the UK Legal Aid Board to find evidence for a planned class-action lawsuit against vaccine manufacturers — a conflict disclosed to no one, while several of the “consecutively referred” children were in fact litigation clients recruited through an anti-MMR campaign rather than ordinary clinical referrals. The retainer predated publication by roughly two years. A study presented as the disinterested observation of a treating physician was, beneath its surface, a brief assembled for paying counsel.
The cost was not abstract. UK MMR uptake fell from roughly 92% in 1995 to about 80% by 2003, dropping below 60% in parts of London — well under the ~95% threshold for herd immunity. Measles, declared eliminated in the UK in the 1990s, returned. The 2012–13 Swansea epidemic produced over 1,200 notified cases, 88 hospitalisations, and the death of a 25-year-old man, Gareth Williams, on 18 April 2013. Wakefield’s claim was not merely unproven; multiple large epidemiological studies covering millions of children — including a Danish cohort of more than 537,000 — found no association whatsoever. This dossier records “Overturned” entry TH-001 as the archetype of the withdrawn: a single twelve-patient paper, amplified by press conference and tabloid, that outran every safeguard in science publishing, peer review, and medical regulation, and was reversed only after a journalist did the work the institutions had not.
For most of the twentieth century the stress-and-acid model was wrong: peptic ulcers were not a verdict on the patient’s worry, ambition or diet but, in roughly 90% of duodenal and up to 80% of gastric cases, a curable infection with Helicobacter pylori. From the Croatian-Austrian surgeon Dragutin (Carl) Schwarz’s 1910 dictum Ohne sauren Magensaft kein peptisches Geschwür — “no acid, no ulcer” — physicians taught that an ulcer reflected too much stress and spice. The promise was a coherent story; the delivered reality was some seventy years of treatments that suppressed acid and managed symptoms while never touching the cause.
The reversal began in Perth, Western Australia. In 1979 pathologist J. Robin Warren saw small curved bacteria colonising the lower stomach of biopsy patients, always alongside inflammation — against the textbook certainty that gastric acid sterilised the stomach. With clinician Barry Marshall, Warren cultured the organism in 1982 (a chance success after an Easter-weekend plate was left incubating past the usual 48 hours), and the pair published in The Lancet in 1983 and 1984. The establishment did not believe them, so in late July 1984 Marshall drank a broth of the bacterium, developed acute gastritis within days, documented it by endoscopy, and cured it with antibiotics — satisfying Koch’s postulates on his own stomach lining.
Displacement was total but slow. A US NIH Consensus panel (12–14 Feb 1994) accepted that H. pylori caused most peptic ulcers and that a short antibiotic course could cure a disease previously managed for life. On 3 October 2005 the Nobel Assembly awarded Warren and Marshall the Prize in Physiology or Medicine “for their discovery of the bacterium Helicobacter pylori and its role in gastritis and peptic ulcer disease” — the formal certificate of a dogma revoked.
This dossier files the stress-and-acid model as TH-002 not as fraud — it was sincere, taught in good faith — but as the family’s purest specimen of an honest, near-universal theory displaced by a confirmed mechanism: institutional confidence, not dishonesty, was the obstacle, and the cost was measured in years of curable suffering prolonged.
On 8 July 2002, the writing group of the Women’s Health Initiative announced that it had stopped the estrogen-plus-progestin arm of its trial roughly three years early, after a mean of 5.2 years, because the combined hormone regimen taken by 16,608 healthy postmenopausal women had increased coronary heart disease rather than preventing it — a hazard ratio of 1.29, alongside elevated stroke (1.41), breast cancer (1.26), and blood clots (roughly doubled). The promise and the reality had been inverted: for nearly two decades, conjugated equine estrogen had been prescribed to protect women’s hearts on the strength of observational data, and the first large randomized test of that promise in healthy women found the hearts were marginally worse off. The gap between the doctrine and the data would be counted in millions of prescriptions written for a benefit that did not exist.
The cardioprotection belief was not a fringe idea. It was mainstream preventive cardiology, taught in medical schools and embedded in clinical guidance, resting on some 40 to 50 observational studies — most prominently the Nurses’ Health Study, which reported current HRT users with coronary heart disease risk reduced by roughly 40 percent (relative risk near 0.61). Conjugated equine estrogen, sold by Wyeth as Premarin since 1942 and as the estrogen-progestin combination Prempro from 1995, exceeded two billion dollars in sales in 2001. The molecules were never recalled; they remain licensed today for menopausal symptoms. What collapsed in July 2002 was the theory attached to them — the claim that they prevented heart disease — displaced not by a scandal or a fraud but by the single most decisive instrument in clinical medicine: a large, randomized, placebo-controlled trial pointed directly at the question the observational data had only ever circled.
The reversal had been foreshadowed. In 1998, the HERS trial — 2,763 women who already had coronary disease, randomized to estrogen-plus-progestin or placebo — had found no overall benefit for secondary prevention, and a 50 percent excess of cardiac events in the first year. The field absorbed HERS as a special case (sick women, late timing) and held to the primary-prevention promise. WHI tested that promise in healthy women, and it failed too. The market answered within months: combined-therapy prescriptions fell by roughly two-thirds within a year, and in 2003 the US Food and Drug Administration imposed a class-wide boxed warning.
This dossier files “Overturned” entry TH-006 as the archetype of a different failure than fraud. No one falsified data. A generation of physicians inferred causation from correlation in self-selected, healthier-than-average hormone users, and only a randomized trial could break the spell — long after the prescription pads had done their work.
In December 2008, four cognitive psychologists — Harold Pashler, Mark McDaniel, Doug Rohrer, and Robert Bjork — published a commissioned review in Psychological Science in the Public Interest that asked the learning-styles industry a single question it had never properly answered: where is the experiment? The “meshing hypothesis” — the claim that pupils learn more when instruction is matched to their preferred sensory style, “visual,” “auditory,” “kinesthetic” — had by then been embedded in teacher training, classroom audits, and commercial inventories across the English-speaking world for roughly three decades. The promise was a tailored, scientific pedagogy. The reality the review documented was a vast literature, thousands of papers, that almost never used the one design capable of testing the claim, and that, in the handful of cases where such a design was used, returned results contradicting it. The gap between the marketing and the evidence was not a shortfall; it was a void.
The required test was specific and unforgiving. Validating styles-based teaching demands a crossover interaction: learners sorted by style, then taught by methods that match or mismatch that style, with the prediction that visual learners do best under visual instruction and auditory learners do best under auditory instruction. Pashler and colleagues found “virtually no evidence” for that pattern. Studies that met the standard tended to show the opposite of what styles theory predicts — that some material is simply better taught one way regardless of who is learning it. The review’s conclusion was that limited education budgets “would better be devoted to adopting other educational practices that have a strong evidence base.”
The debunking did not dislodge the belief. A 2020 systematic review by Philip Newton and Atharva Salvi pooled 37 samples from 18 countries — 15,405 educators — and found that 89.1% still endorsed teaching to learning styles, with national figures above 95% in Turkey, Australia, Greece, South Korea, the Netherlands, and China. The cost is structural rather than catastrophic: teachers asked to produce “four or more versions” of every lesson, finite training hours spent labelling children, and commercial inventories sold into schools on a premise the science had already retired.
This dossier files “Overturned” entry TH-013 as the archetype of the zombie theory: a pedagogical claim with no surviving evidentiary basis, formally refuted in a flagship journal, that continues to be taught as fact because the refutation never reached the people making decisions in the classroom.