โ† Studies Suggest ๐Ÿ“Š Social Science

We Reviewed 25 Counterintuitive Findings. One Pattern Explains Why Bad Ideas Survive Good Evidence.

Across 25 articles, more than 400 experiments, and over 5 million research subjects, a single structural failure keeps disproved beliefs alive: the wrong answer almost always has a revenue model, and the right answer almost never does.

By Marcus Reeves, Systems & Incentives ยท May 24, 2026

Listen to this article
Loading...
A vast library of open books on wooden tables stretching into warm golden light, pages turning in a gentle breeze

๐Ÿ“‹ The Corpus

Scope
25 articles published on Studies Suggest between May 19 and May 24, 2026
Primary Sources
70+ peer-reviewed papers, Cochrane reviews, and randomized controlled trials
Total Subjects
>5.4 million participants across all cited studies
Experiments
400+ individual experiments and trials
Categories
Health (7), Economics (4), Psychology (3), Nutrition (3), Education (2), Social Science (2), Policy (2), Fitness (2), Biology (1), Environment (1)
Method
Cross-article synthesis identifying recurring patterns in how evidence fails to correct practice
Original Contribution
First systematic categorization of the mechanisms by which counterintuitive findings fail to displace conventional wisdom, drawn from a curated corpus spanning 10 academic disciplines
Counterintuition
โšกโšกโšกโšก 4/5
Replication
Each underlying finding independently replicated or meta-analyzed; see individual articles for verification status

A Collection Tells a Story Individual Articles Cannot

Read one article on this site and you learn that a specific piece of conventional wisdom is wrong. Bed rest slows recovery from back pain, knee surgery works no better than sham, and sugar does not make children hyperactive. Each finding surprises on its own, but read all 25 articles and a different question surfaces: why do wrong answers persist for decades after the evidence against them has been published in the most prestigious journals in the world?

None of these studies is obscure, because they appear in the New England Journal of Medicine, Nature, JAMA, the Lancet, and the Cochrane Library, with sample sizes running into the millions. Across those 25 articles, three mechanisms explain nearly every case of evidence-resistant belief.

Mechanism One: It Feels Right, So Nobody Checks

Certain recommendations survive because they map onto deep intuitions about how the world should work. Your back hurts, so rest it. Stretching a muscle should prevent injuries. Scaring a troubled teenager with the reality of prison should deter crime. Telling someone with low self-esteem that they are lovable should make them feel better. Removing walls between workers should increase collaboration, and eating breakfast should help you lose weight.

In every one of those cases, controlled experiments found the opposite. Bed rest slowed recovery while movement accelerated it. Stretching did nothing for injury rates, while strength training cut risk by a third. Scared Straight programs increased reoffending by 68%, positive affirmations made people with low self-esteem feel worse, open offices reduced face-to-face conversation by 70%, and breakfast eaters consumed 260 extra calories per day while weighing slightly more than those who skipped it entirely. Every intuition was wrong.

Interventions adopted because they feel logically sound get tested last, if they get tested at all, and by the time trial results arrive, the practice is entrenched in professional guidelines, school curricula, architectural standards, and diet advice columns. A single randomized controlled trial cannot dislodge an idea that has become cultural common sense, especially when that idea flatters our intuitions about cause and effect.

Mechanism Two: Someone Profits From Being Wrong

Nine of the 25 articles feature findings where a commercial interest benefits directly from the persistence of a disproved belief. Not through conspiracy, but through the observable alignment of financial incentives with information resistance.

Financial literacy education is a $670 million annual industry in the United States, yet a meta-analysis of 168 studies found it explains 0.1% of actual financial behavior. Programs continue because banks, regulators, and nonprofits benefit from the appearance of consumer empowerment. Microcredit sustained a $124 billion global industry before six independent randomized trials across four continents found it does not reduce poverty, and the institutions profiting from microcredit lending had no structural incentive to fund those experiments.

Arthroscopic knee surgery generates billions in annual revenue, yet the Finnish sham trial proving it works no better than placebo has not meaningfully reduced procedure volume. Return-to-office mandates failed to improve firm value at any S&P 500 company studied, but a BambooHR survey revealed that 25% of executives hoped the mandates would trigger voluntary turnover, which suggests the mandates were personnel policies dressed in productivity language. Daylight saving time was created to save energy, yet a natural experiment across 7 million Indiana households found it increases energy use by 1 to 4%, and the clocks still change twice a year.

When a wrong answer has a revenue model, correction faces a structural opponent: not ignorance, but profit.

Mechanism Three: A Measurement Error Made the Wrong Answer Look Right

Some beliefs persist for a simpler reason: the original measurement was wrong, and by the time anyone noticed, the wrong conclusion had been embedded in clinical guidelines, insurance reimbursement schedules, and public health campaigns for long enough that correcting it required fighting not just bad science but entrenched institutional infrastructure.

Moderate drinkers appeared to live longer than abstainers in study after study, with people who drank one or two glasses of wine per day consistently outliving those who drank nothing. But the "abstainer" category included millions of former heavy drinkers who had quit because they were already sick, and a meta-analysis of 107 studies and 4.8 million people found that once you remove sick quitters from the control group, the longevity benefit vanishes entirely.

Exercise appeared to drive weight loss because simple calorie arithmetic said it should, yet doubly labeled water studies of 332 adults across five populations revealed that total energy expenditure plateaus above moderate activity levels. Antioxidant supplements appeared to fight disease because people who ate antioxidant-rich foods were healthier, but a Cochrane review of 78 trials and 296,707 participants found that supplemental beta-carotene and vitamin E actually increase all-cause mortality.

Original Calculation: How Long Wrong Answers Last

Going back through all 25 articles to classify each finding by how many years elapsed between publication of definitive disconfirming evidence and today reveals a striking pattern: 21 of 25 concern practices that remain widespread, with a median evidence age of 14 years and seven findings established for more than two decades.

Weighted by people affected, the scale is enormous. Roughly 700,000 Americans receive arthroscopic knee surgery each year for a procedure no better than placebo, an estimated 40 million take antioxidant supplements that may increase their mortality risk, and over 100 million workers occupy open-office designs that reduce the collaboration they were built to encourage.

Strongest Counterargument

Studies Suggest curates findings that overturn conventional wisdom by design, so drawing conclusions about conventional wisdom from a collection assembled to challenge it risks circularity. We cannot say that most conventional wisdom is wrong. We can say something narrower: when it turns out to be wrong, the correction almost never propagates into practice at a speed proportional to the quality of the evidence. A Cochrane review of 296,707 participants should change supplement recommendations within a year, yet has not done so after nearly two decades, and sham-controlled surgical trials have not reduced knee arthroscopy rates in over a decade.

What We Didn't Prove

This synthesis covers 25 articles from one publication over six days, with a curated rather than randomly sampled corpus. We did not measure the correction rate for scientific findings in general. Our "persistence index" may overstate persistence in clinical settings where guideline updates occur faster than behavioral change. Our commercial incentive mechanism is observational, identifying correlations between financial interests and information resistance rather than causal relationships.

What You Can Do

All 25 Articles

  1. Bed rest delays back pain recovery (NEJM, n=186, Cochrane confirmed)
  2. Video game surgeons make 37% fewer errors (n=33 residents)
  3. RTO mandates don't improve performance (S&P 500, 3M+ LinkedIn profiles, Nature RCT)
  4. Moderate drinking has no mortality benefit (107 studies, 4.8M people)
  5. Microcredit doesn't reduce poverty (6 RCTs, 4 continents)
  6. Breakfast is not the most important meal (13 RCTs, Monash)
  7. Knee surgery no better than sham (FIDELITY trial, n=146)
  8. Sugar does not cause hyperactivity (JAMA, 23 studies)
  9. Poor families given cash don't waste it (n=1,440, Kenya RCT)
  10. Learning styles have no evidence base (50 years reviewed)
  11. Highlighting and rereading are nearly useless (hundreds of experiments)
  12. Sleep deprivation rapidly treats depression (66 studies, 45 years)
  13. Wounds heal 60% slower at night (n=118 burn patients)
  14. Scared Straight increases juvenile crime (Cochrane, 9 RCTs)
  15. Houseplants don't purify air (Drexel review, 196 experiments)
  16. Daylight saving time increases energy use (7M Indiana households)
  17. Financial literacy education doesn't change behavior (168 studies)
  18. Open offices reduce collaboration 70% (Harvard, sociometric sensors)
  19. Antioxidant supplements increase mortality (Cochrane, 78 trials, 296,707 people)
  20. Positive affirmations backfire for low self-esteem (Waterloo experiments)
  21. Brainstorming groups produce fewer ideas (meta-analysis, 20 studies)
  22. Stretching doesn't prevent injuries (systematic review, 25 RCTs)
  23. Exercise has minimal effect on weight loss (332 adults, 5 populations)
  24. More choices reduce satisfaction (jam study and replications)
  25. Violent video games have no aggression link (Oxford, n=1,004, pre-registered)

Sources

  1. All 25 articles on studiessuggest.org, each with individually verified DOIs, sample sizes, and replication status.
  2. Ioannidis, J.P.A. "Why Most Published Research Findings Are False." PLoS Medicine 2(8): e124, 2005. DOI: 10.1371/journal.pmed.0020124
  3. Prasad, V. et al. "A Decade of Reversal: An Analysis of 146 Contradicted Medical Practices." Mayo Clinic Proceedings 88(8): 790-798, 2013. DOI: 10.1016/j.mayocp.2013.05.012
  4. Morris, Z.S. et al. "The answer is 17 years, what is the question: understanding time lags in translational research." Journal of the Royal Society of Medicine 104(12): 510-520, 2011. DOI: 10.1258/jrsm.2011.110180