You open the paper. There is a story about the thing you actually do for a living — your industry, your town, your illness, the software you build. You read it and you wince. The numbers are off. The expert quoted is a nobody. The whole framing is backwards in a way that would take you twenty minutes to explain and the reporter clearly did not have twenty minutes. * Then you turn the page to a story about the Middle East, or interest rates, or a drug trial, and you believe every word of it. * That is the Gell-Mann Amnesia effect, and once you have a name for it you will catch yourself doing it constantly. * **Where the phrase comes from** * The novelist Michael Crichton coined it in a talk called "Why Speculate?", delivered on April 26, 2002 to the International Leadership Forum, a programme run by the Western Behavioral Sciences Institute. His description was simple: you read an article about a subject you know, you see it is wrong end to end, and then you read the next article — on a subject you do not know — as though the paper had suddenly become reliable between page four and page five. * The name is a joke at his own expense, which is worth knowing. Crichton said he picked it because he had once talked the idea over with the physicist Murray Gell-Mann, and attaching a Nobel laureate's name to it made both the effect and himself sound more important than either deserved. It is not a Gell-Mann theory. It is a Crichton observation with borrowed cufflinks. * **Is the underlying premise even true?** * This is the part most people skip, and it is the only part that matters. If stories in your field are wrong at roughly the rate stories in every other field are wrong, then the amnesia is a real error. If your field is unusually badly covered, it is not. * The research says the premise holds up uncomfortably well. Scott R. Maier of the University of Oregon ran the largest modern version of the classic accuracy test — take published stories, send them back to the people quoted in them, and ask what the story got wrong. Across 4,800 articles from 14 newspapers, published in Journalism and Mass Communication Quarterly (Vol. 82, 2005, pp. 533–551), news sources found factual errors in 61% of local news stories and features. * That is not a new collapse, either. Mitchell Charnley ran the first study of this kind in 1936 and found an error rate of 46%. Seven decades of professionalisation, fact-checking desks and style guides, and the number went up rather than down. * Worth saying plainly: sources are graders with a stake. Someone quoted in a story may call an unflattering-but-accurate characterisation an "error." The 61% figure is a ceiling, not a clean measurement. But even generously discounted, it is nowhere near "your field is the unlucky one." * **Why the amnesia survives: the feedback loop is broken** * Here is the mechanism nobody talks about. You would eventually stop trusting the paper if you saw the errors get caught. You do not, because they mostly do not. * Maier followed up in Journalism Practice with a study of 2,700 stories from 10 U.S. metropolitan dailies, tracking what happened after a source spotted an error. Fewer than 2% of the factually flawed articles were ever corrected — about 97% of identified errors went uncorrected. Of the sources who did complain, 130 explicitly asked for a correction; four corrections appeared. * And only around one in ten sources bothered to report the error at all. Maier's interviews found the reasons were not laziness: sources judged small errors not worth the fight, doubted a correction would fix anything, or worried that complaining would get them ignored or punished by the newsroom the next time. * So the loop closes on itself. Errors are common, complaints are rare, corrections are rarer still, and the reader — who never sees a retraction for the piece that annoyed them — is left with no evidence that the pattern extends beyond the one story they happened to be expert in. Amnesia is not stupidity here. It is what happens when the only correction mechanism runs at 2%. If you want to know what the newsroom is actually admitting when it does publish something, our guide to corrections, clarifications, editor's notes and retractions separates the four. * **The trap on the other side** * Crichton's own conclusion was roughly: therefore distrust the whole thing. That is where a lot of people take this, and it is a worse position than the one it replaces. * Blanket distrust is not a higher form of literacy. It is the same lazy move as blanket trust, just wearing a leather jacket — and it is the position that leaves you with no way to tell a wire report of an official statement apart from a columnist's guess about what a foreign leader is thinking. Both get filed under "media lies," and the person who benefits is whoever wants no journalism at all. * The numbers show why the mood is already there: Gallup's October 2025 reading put the share of Americans with a "great deal" or "fair amount" of trust in mass media at 28%, an all-time low, down from 31% the year before and from 68–72% when Gallup started asking in the 1970s. Among Republicans it was 8%. Among Democrats, 51%. A country where trust runs on partisanship rather than on evidence is not a country that has got wiser about errors. * **What to do instead: grade the parts, not the paper** * The useful reflex is not "is this outlet reliable" but "which sentence in this story is load-bearing, and where did it come from." * Separate the sourced from the summarised. "The agency's report lists 4,312 cases" is a checkable claim. "Experts fear the crisis could spiral" is a paraphrase of nobody. In the story about your own field, notice which type the errors clustered in — nine times out of ten it is the connective tissue, not the facts, because the connective tissue is where a non-expert has to guess. * Notice how far the claim travelled. Errors compound with distance from the primary document. A story quoting the study is better than a story quoting a story about the study. Our explainer on primary versus secondary sources is the whole method in one page. * Ask whether anyone stood behind it. A named source with a title can be checked and can be embarrassed. An unnamed one cannot — see our guide to what "according to sources" actually means. * Watch for the press release. A large share of what reads like reporting is a company or agency announcement lightly rewritten, which is why the same phrasing shows up in five outlets on the same morning. That is churnalism, and we have a guide to it. * And when you catch a real error in your own field: report it. One in ten sources does. That is why the correction rate is 2%. * **The short version** * The Gell-Mann Amnesia effect is real, the accuracy research backs the premise, and the reason it persists is a corrections system that almost never fires. But the lesson is not that the news is worthless. It is that "the news" is not one thing to be trusted or distrusted — it is a stack of claims of wildly different quality glued together with paraphrase, and the reader's job is to notice which layer they are standing on. * You already know how to do this. You did it automatically on the story about your own job. The trick is remembering on page five.