Too Good to Check

Everybody knows the forest is wired together. Fungal threads under the soil, trees trading sugar and warnings across a shared network, an old tree in the sunlight feeding a seedling stuck in the shade. It is one of the loveliest facts of the last thirty years, and we put it in three pieces without checking it.

We should not have. In 2023, Karst, Jones and Hoeksema went through the citation record and found the evidence insufficient that these common mycorrhizal networks are widespread in forests, insufficient that transfer through them reliably helps seedlings, and — this is the part that stung — no published evidence at all that mature trees preferentially feed their own offspring. The trade between a root and its fungus is settled science and always was. The forest-wide story is a different claim, and it had been growing by citation rather than by evidence.

We corrected it on 3 August 2026, across No. 5, No. 26, and Field Guide No. 6. And the reason it got past us is not that we were careless. It is that it rhymed with everything we believe. A collection arguing for interdependence and mutual aid met a finding that said the forest is a mutual-aid society, and did not ask it a single hard question.

A story that fits gets less scrutiny, which is exactly when it needs more.

This page is about that failure and its relatives. It is written from inside the problem rather than above it: nearly every example below is one we shipped, or nearly shipped, and had to go back and fix. That is the only honest place to write it from. It is also the reader-facing front door to our fact-check ledger and our public changelog, which have been quietly accumulating this material for a year.

Why the smooth version wins

Misconceptions do not spread because people are stupid. They spread because the retellable version out-competes the accurate one, and the competition is not close.

A paper is hedged, conditional, and boring on purpose. It says in this system, under these conditions, we observed. The retelling drops the conditions, because the conditions are what make it hard to remember. Each hop smooths a little more: the hedge goes, then the sample size, then the species, then the name of the person who found it. What arrives at the other end is a clean sentence that anyone can repeat and nobody can source.

The trap underneath is that repetition feels like verification. When you have met a claim in six places, it has the texture of a well-established fact — but if all six were downstream of the same press release, you have met it once, six times. Our own house rule exists because of this: aggregators and summaries are leads, not citations. A quote site, an encyclopedia, our own glossary, an AI summary — any of them can point you at a source. None of them can be one.

Six ways a fact goes wrong

Sorting a year of our own corrections, they fall into six shapes. The value of a taxonomy over a list of trivia is that you can point it at a claim nobody has written about yet — including one of your own.

01The quote nobody said

The most quotable version of a thought is usually not the one a person actually produced, because people speak in clauses and circumlocutions and the culture files off the corners afterwards. What circulates is a compression — accurate in spirit, unfindable in the text.

“We are star stuff harvesting sunlight” is Sagan's most-shared line, and what is actually wrong with it is more interesting than either the popular version or our own first correction. It is not in Cosmos the book. Searching all thirteen chapters turns up two separate sentences in chapter 9 — one about farming as the methodical harvesting of sunlight, and, elsewhere, “We are made of starstuff.” We concluded the one-liner was a compression nobody had said, marked it as ours in No. 6, and moved on.

That was an overcorrection, and it is the best example on this page. Sagan did say it — aloud, on camera, in the television series: “We are star stuff. Harvesting starlight.” What circulation did was smaller and more specific than invention. It changed starlight to sunlight, ran two sentences into one, and then cited the book, where the line has never been. So the quote is altered, and our correction — having checked only the page and never the broadcast — called a real spoken sentence an invention.

Both errors have the same root: checking one medium and concluding about all of them. A book search cannot settle what a man said out loud. A debunking can be a misconception, which is why that refusal is further down this page and why this entry now carries the correction rather than the tidy version. Caught by Ryan Boren, against Sagan's own voice in the footage, 2026-08-11.

Brian Eno's definition of scenius has the same shape. The version everywhere — including in five places in our own knowledge garden — reads “the intelligence and the intuition of a whole cultural scene… the communal form of the concept of the genius.” That wording entered circulation through Kevin Kelly in 2008. Eno's own sentence, in A Year with Swollen Appendices (1996), does not have those articles in it. A small difference, and it means the line has been quoted as his for eighteen years in a form he did not write.

And sometimes the wrong name is on it. “We are caught in an inescapable network of mutuality” shipped in our Symbioses field guide credited to James Baldwin. It is Martin Luther King Jr., Letter from Birmingham Jail, 1963. Helen Edgar caught it.

02The number that went stale

A measurement can be wrong simply by being old. The literature moves; museum placards, planetarium scripts and popular astronomy do not, and the superseded value keeps circulating for decades with nothing to mark it as expired.

No. 9 built a whole spread on Orion's Belt being an accident of sightline — Alnitak about 740 light-years out, Mintaka 690, Alnilam nearly 2,000, one star three times farther than its neighbours. Those are Hipparcos-era parallaxes, and Hipparcos measured bright, distant, multiple O and B stars badly. Current values from VLTI interferometry and Gaia put all three at essentially one depth, around 1,240–1,260 light-years, all of them in the Orion OB1b association. They are not a coincidence. They are siblings from one cloud.

Two things should have caught it sooner, and both are worth stealing as habits. The zine's own colophon cited an astronomy-picture-of-the-day page that gives the three as a group at about 1,500 light-years and never supported our figures — we cited a source we had not read closely. And Field Guide No. 1, the companion piece, already said it correctly. The collection had been contradicting itself in public for weeks.

03The story told backwards

Sometimes the facts survive the retelling and the emphasis does not, which inverts the meaning while leaving every individual sentence defensible.

Everybody knows everything evolves into a crab. The paper behind that meme — Wolfe, Luque and Bracken-Grissom, 2021 — does say the crab-like body plan has arisen at least five times. It also says it has been lost at least seven times, that the reasons for the repetition “remain a mystery,” and that the authors do not know whether they are looking at true convergence or at deep homology. Three facts that cut against the meme, none of which travel with it. No. 31 is built on the ones that got left behind.

Same shape in No. 32: the shark's cartilage skeleton is routinely called primitive, as though bone had not been invented yet. Minjinia turgenensis, described in 2020, preserves endochondral bone deeper in the tree — so the chondrichthyan condition is a secondary loss. Derived, not primitive. The textbook reading is upside down, and the loaded word turns out to be factually wrong as well as rude.

And in No. 35: the turtle shell is not armour that evolved for protection. Broadened ribs came first as an adaptation for digging, and Lyson and colleagues note they “provide little protection” while costing the animal in locomotion and breathing. The protection is an exaptation — a later use for a structure built for something else.

04The credit that drifted

Attribution decays toward tidiness. A collective becomes a founder; a documenter becomes a coiner; the person with the most citable book absorbs the idea of everyone around them. This one matters more than the others, because it is not only inaccurate — it takes something from people.

The clearest case is in our own field. Neurodiversity was coined by Judy Singer in 1998 is repeated everywhere, including in places that should know better. Botha, Chapman, Giwa Onaiwu, Kapp, Stannard Ashley and Walker published the correction in Autism in 2024: the concept was developed collectively by the online Autistic community, and reached print in 1997 and 1998 through the journalist Harvey Blume, who attributed it to that community. Neurodivergent is Kassiane Asasumasu's, from 2000. The pathology paradigm is Nick Walker's, not Robert Chapman's, who works with the term. Crip time was documented by Zola and Gill, not coined by them. We had two of those wrong in No. 9 and No. 19.

It happens outside our field too, in both directions. Galton did not coin eugenics in Hereditary Genius (1869); he coined it in Inquiries into Human Faculty (1883), in a footnote. “Dirt is matter out of place” is credited to Mary Douglas so reliably that it reads as hers, and it appears unreferenced in Purity and Danger; the earliest instance we could locate is an 1853 address by Lord Palmerston. “Turtles all the way down” is attributed to William James, whose 1882 essay uses rocks; the turtle punchline was attached to him in a 1967 doctoral thesis.

05The press release, not the paper

Between the finding and the public there is a layer whose job is to make the finding interesting, and it is very good at its job. Its output is not a lie; it is a livelier verb.

While building No. 36 we nearly used the story of the bottlenose dolphin that adopted a sperm whale calf. The paper — Wilson and Krause, 2013 — reports “repeated non-agonistic interactions.” “Adopted” is the press's word, not the researchers'. And the explanation everyone repeats, that the dolphin had been harassed by its own kind, is a researcher's suggestion to reporters, not a finding in the paper. The zine says the motive is unknown, because it is.

The costly version of this mode is shark cartilage. A popular book, a market, and decades of belief that sharks don't get cancer — against 42 documented tumours across at least 21 chondrichthyan species, and a clinical trial that found no difference in survival or quality of life. A false premise, industrialised.

06The finding that flatters you

The last one is the wood wide web, and it is the dangerous one, because the others are accidents and this one is an appetite.

Our worst instance was not a fact we imported but an argument we made. No. 33 argued that disgust is learned — a claim that suited us, since the zine is about how revulsion gets recruited into prejudice. Helen Edgar, who taught children with profound and multiple learning disabilities for twenty years, said plainly that this was wrong: pre-term babies have an ick without any capacity to have been taught one. She was right, and Rozin's own framework already carried the distinction we had collapsed. Distaste is chemosensory rejection, present at birth — Steiner's 1973 work found anencephalic newborns produce the gustofacial response, so it cannot be an opinion. Disgust proper is ideational and contamination-sensitive, and that is the only kind our argument was ever about.

We added a spread rather than a footnote, because as published a reader with sensory processing differences could reasonably have understood the zine to be telling them their aversions were prejudice to get over. Aimed at our own community, that would have done harm. And Helen's underlying point is the better framing and now sits in the zine's voice: we are not blank slates — that is the whole point of star stuff. We have history in us. A response conserved across species and present before birth is that history. We had been arguing against this collection's founding premise without noticing.

That card exists as a two-sided printable sheet — the six modes and their tells on the face, the habits and the refusals on the reverse. Print it, pin it up, hand it to someone.

What actually works

None of this needs a laboratory. It needs a handful of stubborn habits, each of which we adopted because something went wrong without it.

• Trace it to the primary. The book, the paper, the essay, the talk. Not a quote site, not an encyclopedia, not our own glossary, not a search summary. Those are leads.
• Cite the source you actually read. The Orion error survived because a colophon named a page that never said what we said. If you have not opened it, it does not go in the references.
• Read your own siblings. Field Guide No. 1 had the belt right while No. 9 had it wrong. Before shipping, check what the collection has already claimed about the same thing.
• When one source falls, re-check its relatives. One 2023 paper on mycorrhizal networks meant re-opening three pieces, not one. Errors travel with their quotes.
• Distinguish a quote from a paraphrase. Verbatim and in quotation marks, or in your own words with the idea attributed. Never dress the second as the first.
• Say what is still open. “This is the leading hypothesis, not a settled result” costs one sentence and is the difference between a claim and an overclaim.
• Log the correction in public, with a date and a name. A living document that never shows a correction is just a document making a claim about itself.
• Measure rather than assume. The same discipline applies to the artifact and not only to its facts — we discovered 44 of our pages printed blank because “prints fine” had been a belief and not a measurement. A stated principle nobody checks is a wish.

Not “do your own research”

There is a version of everything above that curdles, and it is worth naming precisely, because it wears the same clothes. “Go to the primary source” is one short step from “do your own research,” a phrase that has been thoroughly captured and now mostly means trust no one who knows more than you.

The difference is not attitude, it is direction. Tracing a claim moves toward the evidence and ends somewhere checkable, with a citation you can hand to someone else. The captured version moves away from it and ends in a place where nothing can be checked at all, because every institution that could check it has been disqualified in advance. One of them can be wrong and be shown to be wrong. The other cannot, which is the whole appeal.

Debunking has its own failure mode, and it is the mirror image of the one this page is about. Carol Black's essay Science / Fiction describes it exactly: the reverse strategy of asserting “there is scientific consensus where there is no such thing, that an open, ongoing area of scientific inquiry has reached a settled conclusion,” and that anyone disagreeing is “as irrational and ‘anti-science’ as a climate denier.” Her worked example is the wave of LEARNING STYLES ARE A MYTH headlines, which take a narrow experimental result about one simplistic model and inflate it into a verdict on how human beings learn. A debunking can be a misconception. The Myth! headline is as smooth and as retellable as the thing it debunks, which should make you just as suspicious of it.

The refusals

The primary source for a life

Here is why a fact-checking page belongs in a collection about belonging.

The deficit story told about neurodivergent and disabled people has exactly the citation structure described above. It is smooth, endlessly retellable, and repeated until repetition passes for evidence. Its hedges have fallen off. Its credit has drifted onto the people who wrote about us rather than the people who are us. And it flatters the teller, which is why it gets so little scrutiny from the people doing the telling.

Autistic people lack empathy is the wood wide web of our field: a claim that grew by citation, and that Damian Milton's double empathy problem (2012) corrected by going back and looking — the breakdown is mutual, a difference in shared understanding between two parties rather than a deficit in one of them. Neurodiversity was coined by one author in 1998 is the credit that drifted, and it took a 2024 paper by six researchers to put a community's name back on its own idea.

The primary source for an Autistic life is an Autistic person.

That is not a metaphor and it is not a softening. It is the same rule as the rest of this page, applied where it costs something. Miranda Fricker's testimonial injustice names what happens when it is broken — a speaker's account is discounted because of who is speaking — and disability studies has had the shorter version since long before the philosophy arrived: nothing about us without us. Going to the source, here, means asking the person, and then believing the answer enough to let it change what you were going to write.

Which is precisely what happened with the disgust zine. The correction came from a member of our own community with two decades of direct experience, arriving after publication, contradicting the argument we had just made. That is a primary source. We rewrote the piece.

Tending the garden

No. 18 describes this project as a knowledge garden: evergreen, cross-linked, grown rather than filed. Gardens have weeds. The interesting property of a garden is not that nothing unwanted grows in it — it is that someone walks the beds, and a bed nobody walks looks perfectly healthy from the path right up until it isn't.

So the corrections run both ways, which is the cosmic connections point. Checking the scenius quote for No. 31 found the altered wording in five places on stimpunks.org and in none of them the sentence Eno wrote — an error discovered here that has to be fixed upstream. Fact-checking one of Helen's zines surfaced a citation-year discrepancy that neither site can settle alone. This traffic is the part that makes a web of sites worth more than three sites.

Two places hold the record. The fact-check ledger is the working file — every piece, every check, every open item, in the detail needed to re-derive a decision a year later. The changelog is the public version: what we got wrong, what's right, who caught it. Neither is an embarrassment to be managed. They are the only thing that makes the rest of the collection worth believing.

None of this makes us reliable. It makes us checkable, which is the only thing anyone can honestly offer. Read the receipts, follow them back, and tell us where we're wrong — open an issue or write to us. Two of the corrections on this page exist because somebody did.

The forest is not wired together the way we wanted it to be. Fungus and root still trade, nine in ten land plants still depend on the arrangement, and the ordinary version is astonishing on its own. That is usually how it goes. The real thing was rarely the thing that needed improving.