Learn it → Apply it → Save as Skill
When new evidence threatens something we've staked our identity, authority, or past decisions on, the first move is often to attack the evidence, or the person who brought it, rather than to actually weigh it. It's named for a 19th-century physician whose own data could have saved thousands of lives years sooner, if the people who heard it had been willing to update.
Dr. Wallace's Favorite Mental Models is designed to turn useful ways of thinking into reusable tools. Each model gives you a fast visual overview, deeper interactive learning, practical application exercises and an AI-ready Markdown skill you can save for future use.
Instead of merely reading about mental models, you can gradually build your own AI skills library: a collection of portable reasoning skills that can help ChatGPT, Claude, Grok and other AI assistants apply the mental models you value.
The Semmelweis reflex is the reflex to reject new evidence or a new idea specifically because it conflicts with an existing belief, practice, or authority — not because the evidence itself was examined and found weak. The rejection shows up fast, and it usually targets the messenger (their credentials, their tone, their motives) rather than the data.
Important: the second path is not "always believe the new claim." It is examine the claim on its own merits, separately from whether it is comfortable, and separately from who is making you feel defensive by raising it.
The pattern shows up whenever someone (possibly you) hears something that would require admitting a past call was wrong, a long-held practice is harmful, or an expert's standing is less secure than it looked.
Named after Ignaz Semmelweis, a Hungarian physician whose 1847 data on handwashing was rejected by the medical establishment of his day. The pattern he suffered is now used well beyond medicine.
A specific finding, dataset, or observation that contradicts an existing belief, policy, or practice — not a vague feeling, a concrete claim someone could check.
Accepting it would cost something real: a reputation, an authority's standing, a sunk investment, a comfortable identity, or a paradigm a whole field has been trained in.
The response targets the messenger or the claim's plausibility on its face, arriving before the evidence itself has actually been examined on the merits.
No mechanism yet, social cost of being wrong in public, and a community that rewards confident defenders of the status quo over people who raise inconvenient findings.
Answer honestly about your own recent reactions, not how you'd like to react. This stays in your browser; nothing is sent anywhere or scored by AI.
1. When a finding contradicts something I've built my reputation on, my first instinct is to look for flaws in the study, not to ask what would actually change my mind.
2. I've dismissed a critic by focusing on their tone, credentials, or motives rather than the substance of what they said.
3. I feel a flash of irritation or defensiveness before I've actually read the counter-evidence in full.
4. I struggle to recall the last time I changed my mind on something that mattered to me because of new data, rather than just refining how I explain my original position.
5. I've thought or said "that can't be right" about a result before checking how it was actually measured.
6. In my team or group, people who raise inconvenient findings tend to get quietly labeled as difficult rather than thanked.
Open a use case to see what the model changes.
Ask whether the rejection cites a specific methodological flaw, or whether it mainly objects that the result is surprising or inconvenient. Barry Marshall and Robin Warren's H. pylori paper was rejected by The Lancet in 1984 before being published months later once an independent lab replicated it — replication, not reputation, is the right test.
Watch for the team questioning the analyst's methodology before anyone proposes an independent replication. The fix isn't to trust the data blindly; it's to pre-register what evidence would change the roadmap, before looking at results that might threaten it.
Semmelweis's own hospital is the canonical case: a 1847 intervention cut maternal mortality in the First Obstetrical Clinic from roughly 18% to under 2% within months, and it was still largely rejected by his peers for years. Clinical resistance to a protocol change deserves the same question asked of the protocol itself: what's the actual counter-evidence, separate from who is proposing it.
Some delay reflects genuinely needing more evidence or logistics; some reflects institutional reluctance to admit earlier guidance was incomplete. The useful question is which specific objection is being raised, and whether it's falsifiable, not whether the new finding is embarrassing to have been wrong about.
If that person is quietly becoming "difficult" in the team's eyes rather than the data being examined on its merits, the group is closer to Semmelweis's colleagues than it would like to admit. Psychological safety research suggests the fix is structural: reward surfacing inconvenient findings, not just being right.
The most reliable sign is timing: did the objection arrive before you finished reading the evidence? If so, you reacted to the threat, not the claim. Naming that moment, out loud if possible, is usually enough to get back to evaluating the evidence itself.
This is an illustrative example designed to teach the method; the company and numbers are for demonstration only.
Expand for the real history, where the term came from, and an honest look at its limits.
At Vienna General Hospital's First Obstetrical Clinic, staffed by physicians and medical students, maternal deaths from puerperal (childbed) fever ran roughly three times higher than the Second Clinic, staffed by midwives. In April 1847, the First Clinic's death rate was 18.3%. Semmelweis, noting that doctors often went from autopsies straight to delivering babies, required handwashing with a chlorinated lime solution starting that May. By June the rate had fallen to 2.2%, then 1.2% in July and 1.9% in August — roughly a 90% drop, matching the midwives' clinic. He could not yet explain why it worked; germ theory didn't exist until Louis Pasteur's work in the 1860s. His colleagues largely rejected, ignored, or mocked the finding; some resented the implication that a gentleman physician's hands could be unclean. Pushed out of Vienna and later unsuccessful in getting his methods adopted in Budapest, Semmelweis grew increasingly bitter, publishing open letters accusing critics of killing mothers. In 1865 his colleagues had him committed to an asylum, where he died weeks later, at 47, from an infected wound after being beaten by guards. Wide medical acceptance of handwashing came only after Pasteur and Joseph Lister's antiseptic surgery, years after his death.
The phrase is a later coinage, not something Semmelweis himself used. Its exact origin is debated, but the writer Robert Anton Wilson is widely credited with popularizing it, giving it a definition in The Game of Life (co-written with Timothy Leary, first published in 1979). Since then it has been used broadly for reflexive rejection of evidence that conflicts with an accepted paradigm, in and out of medicine — commonly cited alongside the initially resisted ideas of Alfred Wegener (continental drift), and Barry Marshall and Robin Warren (H. pylori causing ulcers, for which they shared the 2005 Nobel Prize in Physiology or Medicine, about two decades after their first rejected paper).
"Semmelweis reflex" is a vivid label, not a precisely defined, independently measured psychological construct the way confirmation bias or authority bias are studied in cognitive-science literature; it overlaps heavily with both of those plus belief perseverance and groupthink, and researchers differ on whether it names something distinct or just restates them with a memorable story. It's also frequently misused as a trump card: calling a critic's objection "just the Semmelweis reflex" can itself be a way to dodge legitimate methodological scrutiny, which is a real risk this page's "When Not to Use It" section addresses directly.
Real skepticism about weak methodology, small samples, or unreplicated results is not this effect. The tell is whether the objection engages the evidence or just notices that the conclusion is unwelcome.
"You're just doing the Semmelweis reflex" can itself be a rhetorical move to avoid answering a real methodological critique. Naming the bias doesn't settle whether the new claim is actually correct.
He couldn't explain why handwashing worked, and that was a legitimate scientific gap at the time, not pure prejudice. Separating "this threatens me" from "this lacks a mechanism" matters for judging any real case fairly.
Plenty of ideas that faced early resistance were, in fact, wrong. Being dismissed is not evidence of correctness; the underlying evidence still has to be checked on its own terms.
The people most confident they'd never do this are often the least likely to notice it in themselves, precisely because noticing requires the same openness the reflex shuts down.
The cases that actually turned out to be right — handwashing, H. pylori, continental drift — won out through independent replication, not through the original claimant simply insisting harder.
the methodology is genuinely weak, the sample is small, or the result hasn't been replicated. Rejecting bad evidence is good practice, not a reflex.
you're tempted to use the label on someone else. It's far more useful as a check on your own reaction speed than as a label to pin on an opponent.
slow adoption is really about cost, infrastructure, regulatory process, or a genuine need for more evidence — not reflexive rejection. Confirmation Bias or Status Quo Bias may fit a given case better.
This browser tool structures your thinking. It does not send your information anywhere or perform AI analysis.
Save as Skill converts the model into portable Markdown instructions that teach an AI when and how to apply the Semmelweis Reflex, what questions to ask, what evidence to seek, what mistakes to avoid, and how to structure its analysis.
Save the skills you find useful and organize them into your own reusable library. The Markdown is intentionally model-neutral, so it can be used wherever an AI system accepts uploaded files, project knowledge, custom instructions or reusable text context. Platform capabilities differ, so this is a portable skill file rather than a claim of one-click installation.
No. Extraordinary claims reasonably need extraordinary evidence, and most new claims that contradict settled findings turn out to be wrong, not revolutionary. The reflex isn't "resistance happened"; it's resistance that targets the claimant instead of the claim, and arrives before the evidence was actually weighed.
Confirmation Bias is the quieter, ongoing habit of noticing and seeking evidence that fits what you already believe. The Semmelweis Reflex is the sharper, more acute reaction to a specific piece of evidence that actively threatens to overturn a belief, decision, or authority right now. They often travel together, but the trigger and the intensity differ.
Germ theory gave the missing mechanism, but adoption still required retraining medical norms, building new institutional habits, and in many places waiting for a generation of physicians trained under the old assumptions to be replaced. A correct mechanism removes the scientific objection; it doesn't automatically remove the organizational and cultural inertia behind the original resistance.