Fossil Reconstruction and Fraud

The forged Archaeoraptor slab is 88 pieces.  An ordinary, honestly prepared Confuciusornis slab, scanned by the same team as a control, is 81 pieces — 22 carrying bone and 55 bone-free shims.¹

"The restored model should therefore be regarded as a working hypothesis, which can be amended with the emergence of new data." — Stephan Lautenschlager, in a methods review of fossil restoration, 2016.²

Body-mass estimates for adult Tyrannosaurus, built from volumetric models of the same skeleton, run from under 6 tonnes to over 18 tonnes — depending on where different researchers put the skin.³

Summary

Piltdown Man.  Nebraska Man.  Archaeoraptor.  These are the standard exhibits in a creationist argument that paleontology cannot be trusted.

That argument is weak, and this article does not make it.

The reason is simple.  In every famous case the fraud was caught by scientists, usually fast, and in the strongest modern case it never passed peer review at all.  Piltdown was exposed by a chemical test that its exposers designed and ran.  Nebraska Man was retracted in Science by the man who co-signed the original identification, after three field seasons mounted specifically to test it.  Archaeoraptor was CT-scanned months before the magazine that announced it appeared, and that scan is what exposed the composite.  A skeptical reader can answer the fraud argument in one sentence, and he would be right to.

The real argument sits right next to it, and it is much better.  It is the gap between the bones that exist and the picture that gets printed.

That gap is not a scandal.  It is a working condition of the field, it is documented by the field itself, and it is measurable.  This article measures it.

The one comparison that makes the point

Archaeoraptor — the smuggled Chinese slab that a US magazine announced in 1999 as a link between dinosaurs and birds — is built from 88 separate pieces:  23 carrying the front half of a bird, 26 arranged to look like an articulated back half, and 39 shims with no bone in them at all.¹

The same team scanned an ordinary Confuciusornis slab as a control.  Nobody has ever accused that specimen of anything.  It is 81 pieces:  22 bone-bearing, and 55 bone-free shims.¹

Read those two numbers together.  A famous forgery and an ordinary, honestly prepared, scientifically usable specimen are built the same way, out of about the same number of fragments, with bone-free shims making up 39 of the 88 pieces in one and 55 of the 81 in the other.  Composite assembly is not fraud.  It is normal preparation.  Fossils from thinly bedded lake sediments shatter when they are dug, the broken edges are too thin to glue, and the standard repair is to mosaic the pieces onto a backing slab.¹

That comparison makes the point without accusing anyone of anything, and it is the argument this article is built on.

What this article concludes

What this article concedes before it argues anything

A skeptical reader should get the bad news first.

Every correction in this article was made by mainstream scientists.  Creationists found none of them.  The Casineria figure error was caught by a European vertebrate paleontologist who travelled to Edinburgh to look at the specimen.  The Brasilodon jaw joint was corrected by the team that micro-CT-scanned it.  The Tikitherium shrew was caught by two Russian mammalogists.  The Spinosaurus reconstruction was attacked by one of its own co-authors.  The shrink-wrapped-dinosaur critique was invented by palaeoartists.  This article is a report on other people's audit work, and it says so.

The underlying objects are real in every case.  The Piltdown bones were a real orang-utan and real medieval humans.  The Archaeoraptor slab was a real bird and a real dinosaur, and both halves were later published as valid species.  The Nebraska Man tooth was a real peccary.  The Casineria skeleton is a real Carboniferous tetrapod.  Nothing here is evidence that fossils are manufactured.

Three arguments in common creationist use die in this article, and they are killed here rather than left for a critic to kill.  The Lucy power-saw video is a misrepresentation, and a creationist with an earth-sciences PhD said so in print in 2026.  Nebraska Man was not at the Scopes trial.  And Piltdown, far from being a keystone that fell, was an outlier whose removal simplified the picture of human evolution — which is what the 1953 paper that exposed it says in its own last paragraph.

No claim here bears on the age of anything.  Every date printed below is the mainstream figure under discussion.  Radiometric dating is not used as evidence anywhere in this series, and a correction to a fossil's classification says nothing about geological time.

What this article rests on

Every quotation was checked against a source that was actually opened, and the ones that could not be checked are named as such.⁴  Where only an abstract could be reached, the footnote says so.

Two numbers in this article are our own.  The Paleobiology Database occurrence counts in Part 6 were pulled live on 3 August 2026, one query per genus, and the queries are printed.⁵  So was the Tikitherium finding in Part 3, which anybody can reproduce in a browser in about thirty seconds.

One quotation that circulates widely in creationist writing on this subject is deliberately left out.  Storrs Olson's 1999 open letter to National Geographic, with its "Piltdown chicken" line, is quoted constantly.  Every copy this project could locate runs through an advocacy website rather than through an archive of the letter itself.  Quoting a secular scientist through a creationist site is exactly the chain-of-custody defect this series exists to remove, so the letter is not used here.

Part 1 — What a fossil looks like before anyone draws it

The slab is not the animal

Start with the physical object, because almost every argument in this article turns on it.

A famous fossil from a place like the Jehol beds of Liaoning, China, or the Solnhofen limestone of Germany, is a slab of thinly layered rock with a flattened skeleton on its surface.  Those rocks split along their bedding planes.  That is why the fossils are so beautifully exposed — and it is also why they arrive in pieces.

Timothy Rowe and colleagues at the University of Texas High-Resolution X-ray Computed Tomography Facility describe the mechanics in an open-access paper, and they are worth quoting exactly because they are describing routine practice, not misconduct.  Thinly bedded fossils shatter on discovery.  The broken edges are too thin to glue back together.  So the standard repair is to lay the fragments out on a backing slab and grout them in place with plaster.  The result, in their words, is "a three-layered object whose 'stratigraphy' is man-made."¹

Two consequences follow, and the paper states both plainly.

First, filler is normal.  Verbatim:  "Surface layers often include extraneous pieces to fill voids around the restored skeleton."¹

Second, so is mixing.  Verbatim:  "It is not uncommon to discover that bones from different specimens, even taxa, were combined to create the impression of a single 'complete' skeleton.  Surfaces are often painted and textured to disguise repairs and give visual continuity to the whole."¹

Their summary of the risk is one sentence:  "A two-century-long history indicates that this type of reassembly is prone to inadvertent error and willful forgery."¹

Note the order of those two nouns.  Error comes first, and it is the bigger category.

The control specimen

Here is the number that should change how a reader thinks about this subject.

Rowe and colleagues scanned a Confuciusornis lacustris slab as a control — a specimen nobody has accused of anything, of the most abundant bird in the Jehol biota.  Their count:  81 separate pieces.  22 of them carry most of the skeleton in correct natural association.  The other 55 are extraneous shims with no bone in them and no verifiable relationship to the skeleton.¹

Do the division:  55 of 81 pieces, or 68%, of an ordinary, honest, publishable fossil slab carry no bone at all.

That is not a scandal and this article does not present it as one.  It is a fact about what the object physically is, and it is the reason the question this article asks — which parts are bone? — is a real question rather than a rhetorical one.

Archaeoraptor, told accurately

Now the forgery, against that background.

The one-paragraph version.  A slab of Early Cretaceous rock was smuggled out of China and sold in the United States for 80,000 dollars.  Whoever prepared it for sale had assembled it from the front half of a fossil bird and the tail of a small feathered dinosaur.  A US popular magazine announced it in November 1999 as a link between dinosaurs and birds, and gave it a Latin name it had no standing to give.  Within about sixteen months the composite had been exposed in Nature.  The tail was published as a real new dinosaur.  The front half was later published as a real new bird.  Both halves were genuine.  Only the arrangement was fake, and the arrangement was made by a preparator working for a market, not by a scientist.

The forensic result.  The CT scan showed the slab's top layer is 88 pieces in three groups.¹

• Group one, 23 pieces, in natural association, carrying the front half of a bird later referred to Yanornis martini.
• Group two, 26 pieces of associated bones arranged to portray an articulated rear half — of which, in the authors' words, "none preserves evidence of a natural attachment to any piece of the first group."
• Group three, 39 shim pieces that "lack bone entirely, or contain only dissociated small fragments of bone."

The scans also recovered the order of assembly from the grout, which is the part of this that reads like forensics rather than paleontology.  Bird skeleton first.  Then a longitudinally split femur.  Then fragmented tibiae and fibulae — "probably both from the left side," and representing the part and counterpart of a single limb.  Then apparent right and left feet, "again, part and counter-part of a single right foot."  Then loose fragments, then the tail, then the shims.¹

Read that carefully.  Two of the three apparent pairs of hind-limb bones are one bone, photographed from both sides of a split slab.  When a rock splits, each bone leaves an impression on both halves.  Lay the two halves side by side and one leg becomes two.

Their overall verdict, verbatim:  "Taken collectively, the Archaeoraptor slab evidently represents two or more species, assembled from at least two, and possibly five, separate specimens."¹

Both halves turned out to be real animals.  The tail was described in 2000 as Microraptor zhaoianus, "the first mature non-avian dinosaur to be found that is smaller than Archaeopteryx."⁶  The front half was published in 2002 under the title "Archaeoraptor's better half," and referred to Yanornis martini, a fish-eating bird.⁷

Two things the popular telling gets wrong

First, it did not fool peer review, because it never went through peer review.

The name "Archaeoraptor liaoningensis" was published in a subscription magazine.  Under the International Code of Zoological Nomenclature that gives it no standing at all — it is not a valid name and never was.  And the CT scanning that exposed the composite was done by Richard Ketcham on 29 July to 4 August 1999, in 422 consecutive slices.¹  National Geographic's article appeared in the November 1999 issue.

The scan came first.  Whatever went wrong here went wrong in a magazine's editorial process and in a fossil market.  It did not go wrong in a journal.

Second, the technical literature is harder on itself than creationists are on it.

Rowe and colleagues note that the Archaeoraptor tail "was reported as part of the type specimen of Microraptor zhaoianus, however, the basis for this assertion was never explained, and this association is no better validated than its prior association with the partial Yanornis skeleton."¹

That is a mainstream team, in a CC-BY paper, saying that the composition of a mainstream type specimen is unverified.  No creationist wrote it.

And one disagreement that is genuinely unresolved

Two mainstream teams disagree about what the Archaeoraptor slab actually is, and nobody has reconciled them.

Zhou, Clarke and Zhang in 2002 say the avialan parts — including hind-limb bones "previously designated as unverifiable 'attached bones'" — come from a single, almost complete individual of Yanornis.⁷  Rowe and colleagues in 2016, fourteen years later, still describe the slab as "at least two, and possibly five, separate specimens."¹

This article prints both and picks neither.  The honest state of the question is that the most-scanned forged fossil in history still has a disputed composition twenty-five years after it was scanned.

What the case actually licenses

Not "scientists faked a fossil."  Nobody in science faked anything here.

What it licenses is narrower.  A fossil that reaches a scientist through a market has been physically assembled by someone whose incentive was to make it look complete, and the assembly is invisible from the surface.  Rowe and colleagues note that several amalgamations and forgeries had passed through their own scanning facility in its first nineteen years, and that the problem grew "as fossiliferous deposits in China began to yield prodigious quantities of fossils."¹  Part 4 takes that up as a live problem rather than a historical one.

Part 2 — Casineria kiddi: the drawing that moved a date for twenty years

This is the best case in the whole subject, and nobody lied.

Nobody forged anything.  Nobody hid anything.  The specimen is real, the workers were competent and respected, the paper passed Nature's review, and the published account was still wrong in a way that moved a first-appearance date for two decades.

The error was a drawing.

What was published

In April 1999, Cheryl Paton, Timothy Smithson and Jennifer Clack described Casineria kiddi in Nature, under the title "An amniote-like skeleton from the Early Carboniferous of Scotland."⁸

The abstract does three separate things at once, and they should be kept apart.

• It names a taxon: "a small, highly ossified, postcranial skeleton of a terrestrially adapted, amniote-like tetrapod from the Mid Viséan; this specimen shows the earliest known pentadactyl manus."
• It reads the arm bone functionally: the humerus is gracile, with a constricted shaft and torsion between its articulations, which the authors treat as "strong evidence of locomotion on land."
• And it infers a ghost lineage — a stretch of evolutionary history with no fossils in it: "Phylogenetic analysis places this new animal close to undisputed amniotes occurring in the Westphalian, indicating that, by the Mid–Late Viséan, amniotes already had a long, but previously unrecorded, history."⁸

Amniotes are the group containing reptiles, birds and mammals — everything that lays a shelled egg or descends from something that did.  Pushing their origin back into the Viséan, and inferring a long unrecorded history before that, is a large claim.

What it rested on

One specimen.

Part and counterpart of a single individual:  NMS G 1993.54.1p and NMS G 1993.54.1cp, in National Museums Scotland, from Cheese Bay in the Gullane Formation of East Lothian.  It is headless.  The skull has never been found.  It is largely tailless as well.  The Paleobiology Database holds exactly one occurrence of the genus, entered from the 1999 paper itself.⁵

Two further facts about that specimen matter, and neither is visible in the abstract.

The wrist is essentially unossified — at most one poorly mineralised element.

And the animal was not finished growing.  The neural arches of its vertebrae stop dorsally at an ossification front capped by a black mineral crust, which is what a bone looks like when it is still being built.⁹  The word "highly ossified" in the 1999 abstract was carrying weight, and the animal it described was a juvenile.

The correction

In 2019, David Marjanović and Michel Laurin published a hundred-page reassessment of Paleozoic limbed-vertebrate phylogeny in PeerJ.⁹  It is not a paper about Casineria.  It is an audit of somebody else's character matrix, and it reports more than 4,200 misscored cells in the original, over half of them typographic.

Casineria is one of the taxa they added.  To add it, Marjanović travelled to Edinburgh and studied the only known specimen for a full day.

Four findings came out of that day.

1.  The claws are not claws, and the published figure is wrong.  Verbatim:  the terminal finger bones "are neither pointed as expected of a claw, nor rounded as incorrectly shown in the interpretative drawing by Paton, Smithson & Clack (1999: fig. 3b).  Instead, their ends are squared off."⁹  Their own figure caption says the same thing more bluntly:  the tip "is blunt, unlike in amniotes and contrary to Paton, Smithson & Clack (1999)."⁹

Claws mattered because claws are an amniote thing.  A small tetrapod with claws is doing something on land that a small tetrapod without them is not.

2.  The vertebrae are not amniote-like either.  The large intercentra resemble those of the anthracosaur Proterogyrinus or the temnospondyl Neldasaurus far more than any amniote's, and they differ from those of Caerorhachis "at most in a slightly lesser degree of ossification."⁹

3.  It may not be a distinct animal at all.  Comparing the specimen against a 2002 redescription of Caerorhachis, they find exactly four differences — Caerorhachis is larger, has more elongate belly scales, a slightly more prominent process on the hip blade, and taller ossified neural spines — and note that all four "can effortlessly be ascribed to ontogeny."  In plain words:  those are the differences between a juvenile and an adult.  Their verdict is that Casineria, once properly redescribed, "could turn out to be either a synonym or the closest known relative of Caerorhachis."⁹

4.  The phylogenetic position collapses.  Their analyses put it "rather far from amniote origins in a gephyrostegid-chroniosuchian grade," and it floats among four alternative positions even there.⁹

The part that makes this a reconstruction story

The claim was already dead in 2006, and the field did not notice for thirteen years.

Jennifer Pawley's 2006 doctoral thesis scored Casineria identically to Caerorhachis apart from missing data, called the two indistinguishable on the available evidence, and recovered Casineria as a temnospondyl — a member of a large group of early amphibian-grade animals nobody would confuse with an amniote.¹⁰

She also stated the general point that guts the 1999 inference.  Quoted inside Marjanović and Laurin's paper:  "none of the postcranial characteristics claimed to be 'reptiliomorph' in Casineria kiddi are truly apomorphic for the amniote lineage.  All are present in temnospondyls […], or potentially may be present in basal temnospondyls (including the five[-]digit manus), because they are plesiomorphic for early tetrapods."¹⁰

Strip the jargon:  every character used to call this animal an early amniote also occurs in animals that are not amniotes, so none of them proves anything.

And her specific prediction about the five-fingered hand was confirmed independently in 2020.  A Polish and German team described an articulated forelimb of the Late Triassic temnospondyl Metoposaurus krasiejowensis with five metacarpals — the first pentadactyl hand known in the group — and concluded that digit number "was not phylogenetically constrained in temnospondyls."¹¹

So the 1999 paper's "earliest known pentadactyl manus" is not evidence of anything amniote-like.  Five fingers is a character that turns on and off.

The twenty-year drift, dated

The most instructive thing about this case is that nobody ever retracted anything.  The claim decayed through the literature instead.

There is no Nature correction.  There is no "Matters Arising."  There is no press release.  The original paper stayed in the citation stream in its original form for twenty years while the field quietly moved the animal four times.

That is more damaging to the popular image of self-correcting science than any fraud story, precisely because nobody behaved badly.  Science did correct itself.  It took twenty years, it never announced that it had, and the correction is buried on page 400-something of a paper about something else.

What this case does and does not allow

It allows this: a first-appearance date for amniotes, and an inferred ghost lineage of tens of millions of years, rested on one headless immature specimen for twenty years — one of its diagnostic characters was an artifact of an interpretive drawing — and the correction came from mainstream paleontologists re-examining the object, published open-access, with creationists nowhere in sight.

It does not allow calling Casineria a hoax, or Paton, Smithson and Clack dishonest.  Nothing in Marjanović and Laurin's paper suggests it.  Clack in particular spent her career deliberately re-collecting the intervals she believed were under-sampled, and the same programme that produced Casineria also produced the fossils that filled Romer's Gap — a famous hole in the record that creationists used to cite and can no longer.¹²

It does not allow the conclusion that amniotes are not old.  A 2025 Nature paper on Australian trackways argues the opposite.¹²  The point here is about how much a picture can carry, not about the age of anything.

One honest limit on this case

The 1999 paper's full text could not be obtained.  Its figure 3b — the drawing with the wrong claws — has not been seen with our own eyes.  Nature serves only the abstract, and no open-access copy could be found.⁸

So the figure error is reported on Marjanović and Laurin's word, from their direct examination of the specimen.  That is a strong source:  they went and looked, they published under their own names in an open-access journal, and they printed the specimen number.  But it is one side of a two-party dispute, and this project could find no reply from Paton, Smithson or Clack.  Absence of a reply is not agreement.  A reader should hold this case at the strength the evidence supports, which is high but not settled.

Part 3 — Two corrections from 2024, and one the database has not made

Both cases below are recent, both were made by mainstream specialists, and both are better arguments than any historical fraud for one reason:  they cannot be dismissed as ancient history.

The anatomy in both belongs to the companion article on the reptile jaw and the mammal ear.¹³  This article owns something narrower — how a published account of a fossil goes wrong, and how long it takes to catch.

3.1 Brasilodon — the bone was real, the inference was wrong

The claim that stood for twenty-one years.  Beginning in 2003, José Bonaparte and colleagues described the small Brazilian cynodont Brasilodon quadrangularis as having an incipient dentary condyle and squamosal glenoid — the beginnings of the mammalian jaw joint.¹⁴

That matters more than it sounds.  The single character that defines a mammal jaw is which bones form the hinge.  Reptiles hinge the jaw between the articular and quadrate bones.  Mammals hinge it between the dentary and the squamosal, and the old articular and quadrate become two of the three bones of the middle ear.  A cynodont with the beginnings of the mammal joint is the closest thing to a mammal that is not one.

So Brasilodon became the sister taxon to Mammaliaformes on exactly the character the whole jaw-to-ear story turns on.  That reading was repeated in the literature for twenty-one years.

The correction.  In 2024, Rawson, Martinelli, Gill, Soares, Schultz and Rayfield published in Nature:  "Brasilodon, contrary to previous descriptions, lacks an incipient dentary condyle and squamosal glenoid and the jaws articulate solely using a plesiomorphic quadrate–articular joint."¹⁵

Why it went wrong is the transferable part, and it is not what a reader expects.

The bone is real.  It exists.  It is the articular process of the dentary, thickened by ridges on its inner and outer faces, extending backwards past the postdentary bones.  Anybody looking at the specimen would see it.

What was wrong was the inference about what it did.  Verbatim:  "The articular process was described previously as an incipient condyle comparable to the condylar process seen in mammaliaforms, but the 3D data show that this thickening is much less developed than in Morganucodon and Sinoconodon, despite the pronounced posterior extension."¹⁵

And the decisive test was itself a reconstruction — a digital one.  Verbatim:  "Digital manipulation of the 3D model of the well-preserved dentaries of Brasilodon UFRGS-PV-1043-T to articulate the jaws shows that the articular process does not reach the squamosal when the teeth are in occlusion, even in the largest specimens."¹⁵

In plain words:  they put the jaw back together in software and closed the mouth, and the two bones that were supposed to be touching did not touch.

That is worth sitting with.  A reconstruction is what created the error, and a better reconstruction is what fixed it.  The difference is that the second one can be published as a dataset, re-run, and checked by anyone.

What made the difference was access to three-dimensional data.  The study micro-CT-scanned 25 specimens — nine Brasilodon, nine Riograndia, seven Oligokyphus — at voxel sizes between 0.0057 and 0.026 mm.¹⁵  The earlier descriptions worked from the outsides of small, partly crushed skulls.  Nobody was careless.  The instrument did not exist.

The honest counterweight, which must be printed.  The same paper's positive finding is that Riograndia does have a clear dentary–squamosal contact, about 17 million years before mammaliaforms have one.¹⁵  So the correction did not delete a transitional feature from the record.  It moved it to a different branch and concluded the feature evolved more than once.

A reader who takes this case as "the mammal jaw transition collapsed" has misread it, and this article says so plainly.  The transferable point is narrower and sharper:  the taxon reported as the closest thing to a mammal was described wrong for two decades, on the one character that mattered, and it took a scanner to find out.

3.2 Tikitherium — and a live query anyone can repeat

The claim.  In 2005 Datta named Tikitherium copei from a single tooth in the Late Triassic Tiki Formation of India, identified it as an upper molar, and made it the oldest mammaliaform in the fossil record — roughly 237 million years old.¹⁶

The correction.  In 2024 Alexander Averianov and Leonid Voyta published in the Journal of Mammalian Evolution under a title that gives away the answer:  "Putative Triassic stem mammal Tikitherium copei is a Neogene shrew."  Verbatim from the abstract:  "We show that this tooth actually represents the last upper premolar (P4) of a Neogene crocidurine soricid insectivore (order Lipotyphla)."¹⁷

So the error was not off by one grade of relationship.  It was an upper molar that is a premolar, of a Triassic stem mammal that is a modern-family shrew, off by something on the order of two hundred million years.

Note which way that cuts before anyone uses it.  An apparently out-of-place fossil — exactly the kind a young-earth writer would be tempted to collect as evidence that the record is disordered — was removed from the record by mainstream mammalogists.  Printing that is the price of using the case at all.

3.3 The database still carries it

Here is a finding this project made itself, by querying the Paleobiology Database directly on 3 August 2026.⁵

• Ask the database for Tikitherium copei. It returns taxon number txn:177642, attribution "Datta 2005," class Mammalia, first and last appearance 237 to 227.3 Ma, interval Carnian.
• Ask it for occurrences of the genus. It returns one: occ:1689327, Tiki Formation, Carnian, class Mammalia.
• Ask it for the reference attached to that occurrence. It returns ref:91378, and the title field reads, in full: "Putative Triassic stem mammal Tikitherium copei is a Neogene shrew".

The database has read the paper that debunks the record, and kept the record.  The 2024 correction is sitting in the occurrence's own reference field while the occurrence still says Mammalia, Carnian, 237 million years.

This is not a complaint about the Paleobiology Database.  Its curators are volunteers, its data are given away free, and it is the single most valuable open resource in paleontology.  This series depends on it and computes original results from it in three other articles.

It is a demonstration of something structural.  A database inherits the literature's errors, and it can inherit them with a lag longer than the correction's own publication.  Any quantitative claim built on a fossil database — including the ones this series makes — is downstream of that.

And the fair thing to do is to check whether the staleness is systemic.  It is not.

The same query set was run against Casineria, whose classification also changed.  The database does not perpetuate the amniote claim there:  it returns parent taxon Tetrapoda, with no amniote assignment.⁵

The staleness is patchy, not systemic.  It is stale where nobody has edited it.  That is a weaker and more accurate claim than "databases are unreliable," and it is the one the evidence supports.

Two further figures from the same live query are worth having.  The database's first-appearance for both Mammalia and Mammaliaformes is 239.48 to 237 million years, in the Ladinian — and that date does not come from either Tiki tooth.  It comes from a Theria sp. occurrence in the German Muschelkalk.  And Gondwanadon tapani — a second single tooth from the same Tiki Formation collection — is still carried as a morganucodontid at Carnian 237 to 227.3 Ma.⁵  So the stale record is not propping up the group's first appearance.  It is simply still sitting there, with the paper that refutes it attached to its own reference field.

What the two cases have in common

Both errors were made by competent people working from the best evidence they had, and both were caught the moment somebody got better evidence — a micro-CT scanner in one case, a mammalogist who knew what a shrew premolar looks like in the other.  Neither is a scandal.  Both are arguments for asking, of any published fossil claim, what physical access the describer had to the object.

Part 4 — Where the specimen came from

This is the strongest current material in this article, and it is better than any historical fraud.

Piltdown is 114 years old and was solved 73 years ago.  A skeptical reader can dismiss it in one sentence.  The commercial fossil trade cannot be dismissed, because it is happening now, it is described in the current technical literature by the people affected by it, and it produces exactly the failure this article is about:  an object that reaches a scientist already assembled by someone whose incentive was to make it look complete.

One caution first.  In every case below, the person who assembled the specimen is a miner, a preparator, or a dealer.  The scientist is downstream.  This is not a claim that paleontologists fake things.

4.1 Spinosaurus — the whole argument in one animal

Spinosaurus aegyptiacus is the sail-backed African predator that has been on magazine covers, in Jurassic Park III, and in a travelling museum exhibit.  It is the most famous non-Tyrannosaurus theropod on Earth.

Everything below comes from mainstream, open-access, peer-reviewed sources, and the fiercest critic of its most famous reconstruction is one of that reconstruction's own co-authors.

The type specimen no longer exists.  Ernst Stromer's holotype, collected from the Bahariya Oasis in Egypt, was destroyed when Munich was bombed in the Second World War.  From an open-access 2015 paper by Serjoscha Evers and colleagues:  "Because the material described by [Stromer] was destroyed during World War II, Stromer's plates and texts remain the only source for comparison."¹⁸

Read that again.  The founding evidence for one of the most famous dinosaurs in the world is a set of drawings in a 1915 monograph.

The replacement was largely purchased.  In 2014, Nizar Ibrahim, Paul Sereno and colleagues published in Science and designated a new specimen, FSAC-KK 11888, as the neotype — the replacement name-bearing specimen.  From it they reconstructed a short-legged, semiaquatic animal.¹⁹  That reconstruction produced the magazine cover and the exhibit.

Evers and colleagues objected, in print, in an open-access journal, and their objections are specific.

• "the allegedly associated skeleton designated as neotype was largely not excavated by the authors, but acquired from a local collector." Parts were bought in 2008 by the University of Casablanca and others in 2009 by the Museo di Storia Naturale in Milan. The collector was located afterwards and led the team to the supposed locality.¹⁸
• "there is no information on the original association of the remains, nor is there additional information on the locality. As most vertebrates in the 'Kem Kem beds' are found in multitaxonomic bone beds, such information is crucial to evaluate if these remains may really pertain to a single individual."¹⁸

And then the sentence that is the reason this case is in this article at all.

The evidence offered that the bones belong to one animal is that the vertebrae match each other in size and shape, and that the left and right limb bones match each other.  Evers and colleagues point out what that argument assumes.  Verbatim:  "local collectors in Morocco and fossil dealers often sort remains acquired from different sources according to fitting size and morphology."¹⁸

The dealer's normal sorting practice manufactures exactly the pattern that is being read as evidence of association.

That is not an accusation of forgery.  Nobody has shown the neotype is a composite, and this article does not claim it is.  It is a statement about what a matching set of bones from a market can and cannot prove.  If a seller routinely groups bones by size and appearance before selling them, then "these bones match" is evidence that they were sold together, not evidence that they lived together.

Their general statement about the source region is in the same paper:  "Moroccan locals collect and excavate material without recording scientifically relevant data.  These fossils are usually purchased and then resold by fossil dealers who operate on a global scale."¹⁸

What happened next, which must be included so this is not read as a hit piece.  Ibrahim and colleagues went back and dug.  In 2020 they published a large part of the tail of the same individual, excavated in place by the team, in Nature.²⁰

And the reconstruction then got harder, not easier.  In 2022 a team built a CT-based skeletal reconstruction with modelled internal air spaces and muscle, and concluded that on land the animal "was bipedal and in deep water was an unstable, slow-surface swimmer (<1 m/s) too buoyant to dive."²¹  Their paper is titled "Spinosaurus is not an aquatic dinosaur."

Its first author, Paul Sereno, is the second author of the 2014 paper he is refuting.

The lesson, stated carefully.  Spinosaurus is not a fraud.  What the case shows is the whole stack this article is about, in one animal:  a destroyed type known only from drawings; a replacement bought from a market that sorts bones by appearance; a reconstruction that became the animal's public image within months; and a competing reconstruction, from the same skeleton and by one of the same authors, that reverses the headline conclusion eight years later.

The dispute about the animal's body plan is live, and this article does not settle it.  Both sides are strong and both are mainstream.

4.2 Myanmar amber — a market that reshaped a research field

Amber from Kachin State in northern Myanmar preserves Cretaceous insects, lizards, feathers and bird wings in three dimensions.  It is scientifically extraordinary.  It is also a commodity.

A 2022 bibliometric study in Communications Biology by Emma Dunne and colleagues counted the publications.²²  The numbers are the point.

The breakpoint at 2014 is not a scientific discovery.  It tracks commercial availability.  Amber mined in Kachin State crosses into Tengchong in China, where it is sold "to jewellers, private collectors, and palaeontologists."²²

The rest of the context belongs in the article too, because it is true and because leaving it out would be selective.  The trade route is one on which the Myanmar armed forces collected taxes after taking the mines in 2017, and the authors cite an estimated annual trade value of one billion US dollars.  Permanent export of fossils from Myanmar is prohibited under its 2015 antiquities law; the amber trade sits in a legal grey zone because amber is also classed as a gemstone.  And their authorship-network analysis shows Myanmar researchers are largely excluded from publications on their own country's material.²²

The lesson for this article is not about ethics.  It is that a research literature can be steered by a market.  The annual publication growth rate rose by a factor of about 45 at the moment the material became purchasable.

A reader who assumes the fossil record is a random sample of the past is wrong twice over.  Once because of what fossilises.  And once because of what gets dug, sold, and shipped.

4.3 Xenodens — a forgery allegation that had to survive testing, and did not

This is the best single example available, because both sides are open access, both are recent, and the test that settled it is one anybody can repeat.

The taxon.  In 2021 Nicholas Longrich, Nathalie Bardet, Anne Schulp and Nour-Eddine Jalil described Xenodens calminechari from one left maxilla — an upper jaw bone — from the latest Cretaceous phosphates of the Oulad Abdoun Basin in Morocco.  Its teeth are small, blade-like, and packed edge to edge like a saw blade, an arrangement the authors say is found in no other squamate or tetrapod.

Where the specimen came from, in the describers' own later words:  it "was acquired unprepared in a block of original matrix in the city of Ouled Bou Ali (Oulad Abdoun Basin, Morocco) from a local farmer."  The fossil comes from an active phosphate strip mine, so visiting the site of the holotype was not possible.  The provenance rests on matrix colour, consistency, associated fish remains, and bone preservation matching a known bed.²³

That is the honest, ordinary situation for a large fraction of Moroccan material, and the describers printed it themselves.

The allegation.  In a paper first published online in December 2024, Sharpe, Powers and Caldwell reassessed the taxon.  From the abstract:  the holotype "was obtained nonscientifically (without technical supervision) from an area in Morocco that yields many manipulated or forged specimens," four tooth crowns appear to occupy two tooth sockets, and "there is potential adhesive connecting the tooth crowns to the maxilla on their lateral sides."  Their conclusion:  "the articulated tooth crowns of this taxon were artificially placed in the maxilla, rendering at least two apomorphies of this taxon the product of forgery," and Xenodens is a nomen dubium — a name of doubtful application.²⁴

Their diagnosis of the general disease is the most quotable line in this whole part.  Verbatim:  "This specimen constitutes a confluence of two persistent problems in vertebrate paleontology: material sourced from commercial excavations that has not been adequately tested for forgery, and taxa named from tooth-based holotypes that ignore the potential for intraspecific dental variation and interspecific convergence in dental characters."²⁴

And they proposed a test rather than just asserting.  They suggested the original authors CT-scan the specimen, and supplied CT examples showing what artificially added tooth crowns look like on Moroccan mosasaur jaws.²⁴

The test was run, and the allegation failed.

In November 2025 Longrich and colleagues published a CT study of the holotype and a second referred maxilla, in the open-access journal Diversity.²³  Their checklist, verbatim:  "Each tooth crown is associated with a single tooth root. … Each tooth root is associated with a replacement pit … Original phosphate sand matrix infills the pulp cavities of teeth and replacement pits.  No evidence of glue, filler or plaster is visible.  Adjacent teeth are similar in size, shape, color, and preservation."²³

And the supposed adhesive turned out to be conservation material.  Verbatim:  the "'gummy material' identified and 'possible adhesive' claimed by Sharpe et al. is shown in cross-section to be rugose, coalesced osteocementum covered with a thin layer of Paraloid B-72 consolidant, as is standard in paleontological preparation and conservation."²³

Paraloid B-72 is the acrylic resin routinely painted onto prepared fossils to stop them crumbling.  A forgery accusation was raised, in a respectable journal, against a surface coating that every prepared fossil in every museum has.

What this case licenses, precisely.  Three things.

1. The commercial-source problem is real enough that mainstream specialists publish forgery allegations against each other's holotypes.
2. It is now testable without damaging the specimen, and the test is cheap enough to be routine.
3. A forgery allegation is itself a claim that has to survive the test. This one did not.

Anyone who cites the allegation without citing the rebuttal is doing exactly what this article accuses others of doing.

One caveat travels with that.  The rebuttal is nine months old at the time of writing, and no reply from Sharpe, Powers and Caldwell could be found.  A reader should check for one.

4.4 Oculudentavis — four months, and why

In March 2020 Nature published a hummingbird-sized skull in Myanmar amber, described as the smallest known avian dinosaur.²⁵

Seven days later, a Chinese team posted a preprint titled "Is Oculudentavis a bird or even archosaur?"  They had re-analysed the original authors' own published CT scan data and found lizard-like morphology.²⁶

The paper was retracted in August 2020.²⁷  In 2021 a better specimen settled it:  "we describe a more complete specimen that demonstrates Oculudentavis is actually a bizarre lizard of uncertain position."²⁸

Four months from a Nature cover claim to a retraction, on a re-analysis of the original authors' own openly deposited scan data.

That is the fastest correction in this article, and the reason it was fast is that the evidence was public in machine-readable form.  Compare Casineria:  twenty years, because checking it required a plane ticket to Edinburgh and a day with the only specimen.

Correction speed tracks data availability, not integrity.  That is the single most useful practical conclusion in this article, and it is one a skeptical reader will agree with immediately.

Part 5 — The picture, and who criticised it first

Credit where it belongs

The best-known criticism of dinosaur art is that it is shrink-wrapped — that artists draw the skin stretched tight over the bones, producing animals that look like flayed corpses rather than living creatures.  A living animal is mostly soft tissue.  Draw a rhinoceros from its skeleton the way dinosaurs are usually drawn and you get something no zoo visitor would recognise.

That criticism was invented and pressed by palaeoartists and palaeontologists.  It is not a creationist insight, and this article will not steal it.

The names are Mark Witton, a palaeontologist and palaeoartist; Darren Naish, a vertebrate palaeontologist; and John Conway, a palaeoartist.  Two of the three co-wrote the 2012 book All Yesterdays, which is where the word "shrink-wrapping" comes from.  In 2014 the three of them published "State of the Palaeoart" in the open-access journal Palaeontologia Electronica.²⁹

If this article presented their argument as a creationist discovery, it would be caught out in one search.

What they actually say

On the state of the work, verbatim:  the current climate promotes and funds "derivative, inaccurate, and sometimes execrable artwork," producing work "that fails to conform with fossil evidence, is incongruous with the anatomy and life appearance of organisms, perpetuates recognised palaeoart clichés and tropes."²⁹

Their definition of what makes a reconstruction inaccurate is a four-item checklist a reader can apply to any museum mural.  Verbatim:  "objectively inaccurate works distort proportions measurable from fossils, omit integumentary structures documented for the taxon concerned, include blatant anatomical errors (regarding, for example, digit number or limb form), or mix geographically and stratigraphically disparate species."²⁹

That last item is worth pausing on.  Putting two animals in one picture is a claim that they lived in the same place at the same time, and it is a claim the picture makes silently.

And their standard for a good reconstruction, verbatim:  it requires "up-to-date knowledge of the taxon being illustrated and the geological context in which it occurs; an ability to reconstruct missing anatomical details (e.g. missing skeletal components and soft tissues) using inference from extinct and modern species; knowledge of those poses and actions which are biomechanically tenable; an understanding of animal colouration and behaviour."²⁹

Three of those four items are inference.  The authors are not hiding it.

A picture-borne error, traced through six publications

This is the best thing in their paper for this article's purposes, because it is a documented chain of copying with named sources, and it is exactly parallel to the Casineria figure error in Part 2.

Robert Bakker drew Barosaurus — a long-necked sauropod — in 1968 and again in 1971.  The drawing has a mast-like neck, a ridge down the midline of the neck, and, because of foreshortening in the original composition, an apparently short tail.

Foreshortening is what happens when something points away from the viewer.  The tail in Bakker's picture is going away from you, so it looks short.  It is not short.  It is a drawing convention.

What happened next, in Witton, Naish and Conway's words:  "Derivative works misinterpreted the latter, thinking the tail of Barosaurus is actually short and introduced other questionable elements which became replicated, such as a horse-like face."²⁹

The chain of images they identify runs through Bakker 1968; Lambert 1978 and 1987; Bartram 1983; Benton 1987; Halstead et al. 1988; and Špinar and Burian 1995 — at least six later publications over twenty-seven years.²⁹

Nobody forged anything.  Nobody even made a scientific claim.  An artifact of perspective in one drawing became an anatomical belief about an animal, and the copyists read the foreshortening as anatomy.

They name other such "palaeoart memes":  the "mega-louse" ankylosaurs, the perpetual scene of a Deinonychus pack attacking a Tenontosaurus, and Ornitholestes chasing a proto-bird.  On the Deinonychus scene they say plainly that the significance of the evidence for that predation "is overstated."²⁹

Why it keeps happening — and the answer is money, not ideology

Their explanation is economic, and this article reports it as such rather than implying a conspiracy of belief.

Publishers and institutions hire in-house "generalist" illustrators to produce derivative works, mainly to save money.  The consequence, verbatim:  "Uncertainty over methods of fossil animal reconstruction and unfamiliarity with the subject matter forces 'generalist' artists to base their work closely on that of dedicated palaeoartists, and changes made to the original work — even minor ones — typically introduce errors.  Conversely, outright copying means that mistakes or superseded components of the original artwork are perpetuated rather than corrected."²⁹

Their own analogy is a painting by John Conway of a modern cat drawn with scales.  If an artist depicted a living animal that badly, they write, "their work would be considered an abject failure and future commissions unlikely.  Fossil animals are frequently depicted with equivalent gross inaccuracies however."²⁹

And they distribute the blame in both directions, which keeps this from being an anti-science point.  "some of the largest and most respected scientific institutions and publishers engage in these unethical, copyright-infringing acts."  And equally:  "some palaeontologists employed as consultants consider artistic depictions of extinct animals unimportant or unworthy of checking, correcting or replacing over time."²⁹

One thing this article cannot verify.  All Yesterdays is reported to contain a reverse experiment — reconstructing familiar modern animals from their skeletons alone, using the conventions of palaeoart, to show what those conventions do to animals whose appearance is already known.  The book was not obtained for this article, so that experiment is described here only as a report, and no detail from it is printed.³⁰

How much of the picture is actually fixed by bone? A 2025 number

The shrink-wrap criticism is qualitative.  There is now a quantitative version, and it is the strongest single number in this article.

In 2025 Dempsey, Cross, Maidment, Hutchinson and Bates published a modelling study of 52 non-avian dinosaurs in Biological Reviews.³  It is not a polemic.  It is a methods paper about how body size and shape are estimated.

Their statement of the problem, verbatim from the abstract:  "the determination of body volumes in fossil taxa is often subjective, and may be sensitive to varied artistic inference."³

And in the body:  "Based on the observation of skeletons alone, the biologically feasible minimal and maximal extents to which soft tissue outlines should adhere to skeletal landmarks are unavoidably subjective, and results may vary among investigators."³

The number.  Verbatim:  "Depending on the size and shape of the soft tissue envelope estimated by different researchers, body mass estimates based on volumetric models of adult Tyrannosaurus (~11–12 m in length) range from less than 6 tonnes to over 18 tonnes."³

A threefold range in the body mass of the best-studied dinosaur on Earth, from the same skeleton, entirely because of where different people put the skin.

That is from a 2025 review in a mainstream journal, not from an apologist.

Supporting figures from the same paper, all directly quotable:³

• Their own tail-volume estimates carry a mean percentage prediction error of ±52%. The torso — the largest and best-constrained segment — carries ±10% or ±14%, depending on which modern group supplies the equations.
• Total body-mass error bars run from 74–86% of the point estimate at the low end to 114–126% at the high end.
• Choosing minimum versus maximum uniform bird-like densities alone shifts the modelled mass between 79–86% and 125–137% of the model that varies density by body region.
• A widely used protocol derived from CT scans of modern mammal skeletons found that minimal convex hulls "consistently underestimated body mass by approximately 21%."

And the direction of the correction runs against the shrink-wrapped tradition:  "we infer that many previous reconstructions of soft tissue envelopes may be too small, and that many dinosaurs were therefore heavier than previous estimates."³

Now the honest limit, which matters more than the headline.  This is not "dinosaur masses are unknowable."  The same authors rank all 52 taxa by mass and find the ordering holds across methods at a Spearman correlation above 0.99.³  Whether Tyrannosaurus was heavier than Allosaurus is robust.  Whether it weighed six tonnes or eighteen is not.

Robust relative claim, soft absolute claim.  That distinction is precisely what this article is trying to teach, and it applies to creationist arguments as readily as to anyone's.

The rule the field states about its own restorations

In 2016 Stephan Lautenschlager published a methods review of digital fossil restoration in Royal Society Open Science — remeshing broken surfaces, mirroring damaged elements, duplicating vertebrae, undoing tectonic distortion, extrapolating missing regions.²

His warnings escalate in a specific order, and they are worth having in that order.

• On duplicating elements: these operations "are inherently more subjective and require greater interpretation."²
• On undoing distortion: it "might not only introduce a large degree of interpretation and uncertainty, but could also potentially alter the biological signal contained in the fossil."²
• On the whole workflow: "each restoration step not only increases in complexity, but also in the degree of interpretation introduced into the restored fossil. This subjectivity is often further amplified by the necessity to perform these restorative steps manually, precluding easy and unbiased reproducibility and transparency of the restoration process."²
• On the fix: "A detailed documentation of the restoration steps should therefore accompany works using this approach."²

He also names the trap that makes single-specimen taxa dangerous, which is where Part 2 and Part 6 meet.  Filling a gap from another specimen of the same species "is often not possible in palaeontology, where taxa are frequently represented by a single specimen only," and borrowing from a related species "can be highly problematic as it assumes an absence of morphological variability across different taxonomic units."²

And his conclusion is the sentence this whole article is arguing for, written by a mainstream paleontologist in an open-access journal:  "The restored model should therefore be regarded as a working hypothesis, which can be amended with the emergence of new data."²

That is not skepticism.  It is a request for a caption.

Part 6 — How much material famous taxa actually rest on

Hominins are deliberately excluded from this part.  The companion article on pre-Homo hominins owns them, and including them here would let a reader dismiss the whole table as an argument about human origins.³¹  Every animal below is a non-hominin whose picture most readers have seen.

What the numbers are and are not

Two different kinds of count appear in the table, and confusing them would be a mistake.

Paleobiology Database counts were pulled fresh for this article on 3 August 2026, one query per genus.⁵  An occurrence is one taxon recorded in one collection.  A collection is one sampling event at one place and horizon.  Neither is a specimen count.  A single collection can hold two hundred bones or one tooth.  These numbers are a lower bound on how many separate places a genus is known from, and nothing more.  They also inherit the literature's errors — see Part 3, where the database still carries a Neogene shrew as a Triassic mammal.  Use them for order of magnitude, never for a precise claim.

Specimen facts in the right-hand column were each read out of a named source, and the source is cited where the claim is used.

The table

The pattern is not "everything is one bone."  Basilosaurus and Confuciusornis are richly sampled.  Two rows in an earlier draft of this table — Therizinosaurus and Megalosaurus — were cut because their material was not audited, and a table like this is worthless if any row in it is a guess.

The real pattern is narrower and more interesting.  The taxa that carry the most argumentative weight are often the thinnest.  A taxon becomes famous by being the earliest, or the most transitional, or the largest.  Being the earliest usually means being known from one place, because the earliest record of anything is by definition the one nobody has beaten.

The extreme case: a species named, and reclassified 140 years later, from a drawing

In 1878 Oramel Lucas shipped Edward Drinker Cope a huge fossil bone from Colorado — a neural spine from a back vertebra, 1.5 metres tall.  Cope named the animal Amphicoelias fragillimus, published a description and an engraving of the bone, and the specimen was never seen again.

Cope's collections were bought by the American Museum for 60,550 dollars after his death in 1897 — about 1.8 million dollars in 2018 money — and the transfers were not complete until 1903.  Missing specimens, this one among them, were assigned catalogue numbers anyway in case they ever turned up.  They did not.³²

In 2018 Kenneth Carpenter reassessed the animal and moved it to a new genus and a different family.  From his abstract, verbatim:  "The holotype was lost and all that is known of the specimen is from Cope's original publication."³²

And the rules of zoological naming permit exactly this.  Carpenter quotes Article 72.5.6 of the International Code of Zoological Nomenclature:  "In the case of a nominal species-group taxon based on an illustration or description … the name-bearing type is the specimen or specimens illustrated or described (and not the illustration or description itself)."³²

So an animal sometimes claimed as the largest that ever lived is known from one bone that no longer exists, and in 2018 it was moved to a different family on the evidence of an 1878 engraving.

That is a true statement, it is sourced entirely to a mainstream open-access paper, and it is the cleanest single illustration of this article's thesis.  It is also not an argument against sauropods, against the Morrison Formation, or against anything else.  It is an argument about what a picture can carry.

Part 7 — The famous frauds, told accurately

These are the cases every reader expects.  They are last rather than first because they are the weakest material here, and because told accurately they do not support the argument they are usually cited for.

7.1 Piltdown (1912–1953)

What was announced.  On 18 December 1912, Charles Dawson — a Sussex solicitor and amateur antiquarian — and Arthur Smith Woodward, Keeper of Geology at the British Museum, presented to the Geological Society of London an ape-like jaw with two worn molars, plus parts of a human-like braincase, said to come from a gravel pit at Piltdown in Sussex.  Stone tools and fragmentary fossil mammals came with them, all stained dark reddish-brown to match the gravel.  Digging in 1913–14 added a canine tooth intermediate in size between an ape's and a human's, and a carved bone slab immediately nicknamed the "cricket bat."  In 1915 Dawson reported a second site about 3 km away, whose exact location he never told anyone.  He died in August 1916, and nothing significant was ever found at either locality again.³³

What it really was.  A modern orang-utan jaw and canine, filed down, stained and planted; parts of a modern — possibly medieval — human skull; and genuine fossil mammals gathered from other places.  The "cricket bat" had been carved with a steel knife.

How it was exposed, which is the part that matters

The 1953 paper is titled "The Solution of the Piltdown Problem," by Joseph Weiner, Kenneth Oakley and Wilfrid Le Gros Clark.³⁴  Its structure shows that the exposure was a test, not a rumour.  Four independent lines of evidence.

1.  Direct observation of artificial abrasion.  Eight numbered points on the molars, four on the canine.  Verbatim, on the molars:  "The borders of the flat occlusal surfaces particularly the lateral borders are sharp-cut and show no evidence of the bevelling which is usually produced by natural wear."  And:  "Inspection of the isolated molar tooth … with a binocular microscope shows that the occlusal surface of the enamel has been finely scratched, as though by an abrasive."³⁴

2.  Fluorine.  Buried bone absorbs fluorine from groundwater over time, so two bones from the same bed should carry similar amounts.  They do not.  The Piltdown jaw is less fluorinated than a modern chimpanzee's tooth.³⁴

3.  Nitrogen.  Fresh bone runs about 4.1% nitrogen and loses it with age.  Upper Pleistocene bone from London gives 0.7%, and the Piltdown skull fragments 1.4%.  The Piltdown jaw gives 3.9%.³⁴  The paper heads off the obvious objection in its own parentheses:  if the specimens had been soaked in a preservative, "the cranial bones which are porous would have shown more nitrogen than the highly compact dentine of the teeth whereas the reverse is true."³⁴

4.  Staining.  The canine's black coating turned out to be "a tough, flexible paint-like substance, insoluble in the common organic solvents."  The skull fragments were iron-stained through their whole thickness; the jaw's staining was superficial.  And there is the chromate.  Smith Woodward had recorded that Dawson dipped the first skull pieces in potassium bichromate.  Pieces dug later, in Smith Woodward's presence, contain no chromate.  The jaw does.³⁴

Their conclusion, verbatim:  "The iron and chromate staining of the Piltdown jaw seems to us to be explicable only as a necessary part of the deliberate matching of the jaw of a modern ape with the mineralized cranial fragments."³⁴

And note who ran the experiment.  The trigger for the whole investigation was Weiner putting the fraud hypothesis "fairly and squarely as the only possible solution," pointing out that nobody had ever examined the organic content of the jaw — and then filing and staining a chimpanzee's teeth himself, producing "an appearance astonishingly similar to the Piltdown molars and canine."³⁴

A creationist reader should notice which side did that.

The 2016 study

In 2016 a sixteen-author team led by Isabelle De Groote, with Chris Stringer last, published new work in Royal Society Open Science under the title "New genetic and morphological evidence suggests a single hoaxer created 'Piltdown man'."³³

• Ancient DNA. 127 base pairs of mitochondrial DNA were recovered from the Piltdown I canine and the Piltdown II molar. Both carry the same haplotype, consistent with one animal, and both place in the Bornean orang-utan Pongo pygmaeus rather than the Sumatran species. DNA from the human skull fragments failed entirely, so the humans could not be sourced.³³
• A single method of working. Gravel from the Piltdown site was packed inside the tooth cavities and skull cavities of specimens from both sites, held in with small pebble plugs. The same zinc- and zirconium-rich white putty, resembling a dental filling material, was used throughout.³³
• Damage nobody had reported before. Micro-CT shows the jaw was hollowed out beneath the molars and cracked through the tooth sockets, most likely by pulling the jaw apart at the chin to widen the sockets and get the teeth out for filing. The forger reseated two teeth in putty and left the third socket empty.³³
• The verdict. One orang-utan for all the ape material at both sites, and "two or perhaps three, possibly medieval, humans." Verbatim: "The consistency in the MO observed in the specimens, and the use of a limited number of specimens to create both the Piltdown I and Piltdown II material, are indicative of a single forger. This was most likely Charles Dawson—the prime suspect since the fraud was exposed in 1953."³³
• And the forger was an amateur. Verbatim: "our novel analyses of the materials from Piltdown have shown that the forger was not a trained conservator. Some aspects of the work show inexpert skills, resulting in bones fracturing, putty setting too fast and teeth cracking while being filed down."³³

The part that runs against the apologetic use of the case

Piltdown never fitted, and specialists said so from the start.  De Groote and colleagues:  "There were early doubters who suggested that a more ancient ape fossil could have become mixed into deposits containing a more recent human cranium."³³

It was already being sidelined before it was exposed.  Verbatim:  "During the course of the 1920s and 1930s, Eoanthropus became increasingly marginalized as ancient hominin fossils were discovered in China, Indonesia and Africa, none of which showed the peculiar combination of an ape-like jaw and human-like braincase found in 'Piltdown man'."³³

And the 1953 authors said the same thing in their last paragraph.  This is the single most important quotation in the Piltdown file, and it cuts against our own side.  Verbatim:

the elimination of the Piltdown jaw and teeth from any further consideration clarifies very considerably the problem of human evolution.  For it has to be realized that 'Piltdown Man' (Eoanthropus) was actually a most awkward and perplexing element in the fossil record of the Hominidae, being entirely out of conformity both in its strange mixture of morphological characters and its time sequence with all the palaeontological evidence of human evolution available from other parts of the world.³⁴

Piltdown was not a keystone that fell.  It was an outlier whose removal simplified the picture.  Any argument that treats it as evidence that the rest of the record is a house of cards has the logic exactly backwards, and a hostile reader will find that paragraph in ten minutes.

Two lessons that do transfer

Confirmation bias was a real cause.  De Groote and colleagues note that Dawson "would have known what the British scientific community was anticipating in a missing link between apes and humans: a large brain, an ape-like face and jaws."  Their own moral, verbatim:  "it stands as a cautionary tale to scientists not to see what they want to see, but to remain objective and to subject even their own findings to the strongest scientific scrutiny."³³

Restricted access to the originals is what let it run for forty-one years.  Casts were circulated while "the originals were generally kept locked away and only shown to a select few."  And:  "Perhaps, if the five years of World War I had not intervened, and more people had been given access to the original material … the forgery would have been discovered earlier."³³

That is the transferable lesson, and it is the same one as Part 4's.  Access to the object is what catches errors.  It is not "scientists are dishonest."

And there is a reconstruction story buried inside the fraud story.  De Groote and colleagues record that Arthur Keith publicly ridiculed Smith Woodward's skull reconstruction at the International Medical Congress on 11 August 1913, as reported by The Times the next day — and that the canine tooth, "remarkably similar to the model one Smith Woodward had inserted in the reconstructed jaw," turned up eighteen days later, on 30 August, silencing the critics.³³

A model tooth in a reconstruction predicted the fake that was then produced to match it.

The fair criticism to concede.  Forty-one years passed between announcement and exposure.  During that time Piltdown occupied textbook space, museum space, and one written expert statement at the Scopes trial.  The reason it lasted is not creditable:  it flattered a national expectation and a brain-first model of human evolution, and the institution holding the originals restricted access to them.  Say that plainly, and stop there.

7.2 Nebraska Man (1922–1927) — where our own side's telling fails

This is the case where the popular creationist account does not survive a check against the primary literature.  Everything below was read in the original.

What Osborn actually described.  Henry Fairfield Osborn published "Hesperopithecus, the First Anthropoid Primate Found in America" in American Museum Novitates on 25 April 1922.³⁵  The material was one water-worn tooth, crown 10.5 by 11 mm, found by Harold J. Cook, a consulting geologist, in the Snake Creek beds of Nebraska.

The sentence creationists quote is genuine.  Osborn printed his own reaction:  "The instant your package arrived, I sat down with the tooth, in my window, and I said to myself: 'It looks one hundred per cent anthropoid.'"³⁵

But it is Osborn quoting his own letter of 14 March 1922, written before study, in a paper that then hedges hard.  His actual taxonomic claim, in the same publication, verbatim:  "it would be misleading to speak of this Hesperopithecus at present as an anthropoid ape; it is a new and independent type of Primate, and we must seek more material before we can determine its relationships."³⁵

And in Nature, four months later, he refused to call it an ape-man at all.  Osborn's August 1922 Nature paper was written to rebut both an over-enthusiastic friend and a skeptic.³⁶  Its last paragraph states the exact point this article is making, in 1922, in Nature:

I have not stated that Hesperopithecus was either an Ape-man or in the direct line of human ancestry, because I consider it quite possible that we may discover anthropoid apes (Simiidæ) with teeth closely imitating those of man (Hominidæ) … There are so many crisscross adaptations of this kind among the mammals that we can never be sure about the family relationships of an animal until we secure not only the teeth but considerable parts of the skeleton as well.³⁶

Who doubted it, and how fast.  Arthur Smith Woodward — the Piltdown man himself — published an objection in Nature on 10 June 1922, under seven weeks after the description, proposing comparison with a Pliocene bear instead.³⁷  William King Gregory later recorded that "nine suggestions were put forward by responsible critics as to what the type specimen of 'Hesperopithecus' might represent other than any kind of ape or man."³⁸

The American Museum's own two specialists split in print in 1923.  Verbatim, as Gregory quotes their joint statement:  "one of us (M. H.) still regards the human resemblances as being of considerable significance, while the other (W. K. G.) leans toward the anthropoid affinities of the type."³⁸

And a warning already existed in the museum's own literature.  In 1909, W. D. Matthew and Harold Cook — the same two men who later handled the tooth — wrote in a Bulletin of the American Museum, verbatim:

The anterior molars and premolars of this genus of peccaries show a startling resemblance to the teeth of Anthropoidea, and might well be mistaken for them by anyone not familiar with the dentition of Miocene peccaries.³⁹

Print the next sentence too, because it is routinely dropped.  Matthew and Cook continue that the improbability of finding anthropoid teeth in this formation "is substantially decreased by the discovery of true antelopes at this horizon."  They were arguing that an Asian-origin fauna made a stray anthropoid less improbable — not warning anyone off the idea.³⁹

The retraction.  Osborn sent Albert Thomson back to the Snake Creek beds in the summers of 1925 and 1926.  On 16 December 1927 Gregory published in Science under the title "Hesperopithecus Apparently Not an Ape Nor a Man."  Verbatim:  the expedition "secured a series of specimens which have led the writer to doubt his former identification of the type as the upper molar of an extinct primate, and to suspect that the type specimen of Hesperopithecus haroldcookii may be an upper premolar of a species of Prosthennops, an extinct genus related to the modern peccaries."³⁸

Gregory prints the remaining weakness in his own case rather than hiding it:  in Prosthennops the comparable premolars have two inner roots, and the type tooth has one broad root.³⁸

The postscript, added after the 1927 field season, is the end of it:  "every one of them appears to me to pertain to Prosthennops … Thus it seems to me far more probable that we were formerly deceived by the resemblances of the much worn type to equally worn chimpanzee molars than that the type is really a unique token of the presence of anthropoids in North America."³⁸

Five years and eight months from description to retraction, in Science, by the man who co-signed the original identification, on the strength of three field seasons mounted to test it.

The picture, and the Scopes claim

The famous drawing was an artist's, not a scientist's.  A. Forestier's two-page spread of a pair of very human-looking Hesperopithecus individuals ran in The Illustrated London News on 24 June 1922, alongside an enthusiastic article by G. Elliot Smith.⁴⁰  Forestier's own stated method was to model the animal on Pithecanthropus, "the Java ape-man, whose proportions and attitude were those of the average Englishman."

Did Osborn endorse it?  The line usually quoted is that Osborn and his museum colleagues thought such a drawing "would doubtless be only a figment of the imagination of no scientific value, and undoubtedly inaccurate."  That is real, but it is not a signed statement by Osborn.  It comes from an unsigned New York Times article of 17 September 1922 reporting the museum's view, and this project could not obtain that newspaper page.⁴¹

So the honest position is:  there is no evidence Osborn endorsed the reconstruction, and the only evidence of any reaction says the museum dismissed it — reported at second hand, and unverified here.  Never attribute the "figment of the imagination" wording to Osborn's pen.

Was it used at the Scopes trial?  No, and this was checked directly.

Two independent full-text copies of the complete stenographic report of the 1925 trial were downloaded and searched.⁴²  The results:

Neither the tooth, nor its name, nor Osborn's name appears anywhere in the trial record.  Osborn had been named to the defence's expert list, but he never submitted written testimony and never went to Dayton.

What is on the Scopes record is Piltdown — flagged as disputed, by the defence's own expert.  Horatio Hackett Newman's written statement, filed after the judge excluded live expert testimony, says of Eoanthropus:  "Until the experts arrive at an agreement about this type it might be well for others to reserve judgment."⁴²

And Newman's statement contains, unprompted, the best 1925 statement of this article's own theme.  On Pithecanthropus, verbatim:  "There is no proof that these remains belong to the same individual or even to the same species… Many pages of scientific romance have been written about this species; all sorts of more or less justifiable pictures and models of this hypothetical species have been published."⁴²

"Scientific romance," and "pictures and models of this hypothetical species," is a defence witness at the Scopes trial.  Not an apologist.

Where Osborn is genuinely open to criticism

He used the tooth polemically against William Jennings Bryan, and he did it in a book.  In The Earth Speaks to Bryan (1925), verbatim:

Certainly we shall not banish this bit of truth because it does not fit in with our preconceived notions and because at present it constitutes infinitesimal but irrefutable evidence that the man-apes wandered over from Asia into North America.⁴³

"Infinitesimal but irrefutable," of a single worn peccary premolar, in a book written to humiliate a religious opponent, at a moment when his own two co-workers had already said in print that the diagnosis must await further discoveries.

That is a real fault, and this article says so.  It is the strongest true thing in the creationist file on this episode, and it is not the thing creationists usually say.

Summary verdict on the popular account

Part 8 — The Lucy power-saw clip, killed here

This section exists to stop this site publishing a false claim.

The claim is that Owen Lovejoy took a power saw to Lucy's fossil pelvis to make it look human.  It is the most-circulated creationist argument about reconstruction.  It is wrong on the fact that matters, and anyone who has watched the clip has already heard the sentence that refutes it.

The full anatomical question — whether Australopithecus afarensis walked upright, and how it should be classified — belongs to the companion article on pre-Homo hominins.³¹  This article owns the media artefact.

What is in the clip

The video is 52.8 seconds, excerpted from the NOVA documentary "In Search of Human Origins," Part 1, broadcast by WGBH and PBS in 1994.  Creationist reposts circulate it under titles like "Fixing Lucy's Pelvis with a Power Saw."

The tool is real.  Frames extracted from the clip show a hand-held motorised rotary cutter with a small disc, throwing dust, cutting a pale fragment held between the fingers, at roughly 16 to 19 seconds in.  The popular title is visually accurate about the tool.

And the narration in the same 52.8 seconds says the object being cut is a copy.  The PBS transcript is free and online.⁴⁴  Lovejoy, on camera, explains that the crushed part of the pelvis had broken into individual pieces which then fused during fossilisation, and that this made the two bones "fit together so well that they're in an anatomically impossible position."  Donald Johanson's narration then says, verbatim:

He didn't want to tamper with the original, so he made a copy in plaster.  He cut the damaged pieces out and put them back together the way they were before Lucy died.⁴⁴

Both facts are in one 53-second clip.  That is not a subtle context problem.  It is quoting the picture and cutting the sound.

The creationist who says so in print

Christopher Rupe holds a PhD in the earth sciences and is a young-earth creationist.  In 2026 he published a twenty-page paper in the Creation Research Society Quarterly arguing that Lucy's kind should be reclassified — a position friendly to this site's own leanings.  In the middle of it he writes, verbatim:  "The allegations of fraud or a manipulated reconstruction do not stand up to scrutiny."⁴⁵⁴⁵

His reason is anatomical and checkable.  Quoting Lovejoy's own description of the damage — "While much of the iliac blade is well preserved, the posterior third has been crushed, crumpled, and bent anterolaterally almost 90°" — Rupe points out that the crushed part carries the joint face that meets the sacrum.  Fit the damaged hip bone to the sacrum as found, and the whole thing rotates until the lower parts of the pelvis point straight out sideways, which he calls an anatomical impossibility.  It is only in that impossible position that the blades lie flat like a chimpanzee's.⁴⁵

On Rupe's reading, the un-reconstructed, chimp-looking Lucy that creationist audiences are shown is the artefact, and Lovejoy's restoration is the correction.

And the reconstruction can be taken out of the argument entirely

Rupe's stronger move is to show that the answer does not depend on Lovejoy's work at all.  Four independent lines reach the same shape.⁴⁵

1. Lovejoy never touched the sacrum. It was found intact, with distortion concentrated on one side, correctable by mirroring.
2. Peter Schmid in 1983, and Schmid with Martin Häusler in 1995, produced independent reconstructions. Rupe stacks all three in one figure. The disagreements are about birth-canal dimensions for obstetric analysis, not about whether the thing is bowl-shaped.
3. A 2018 volumetric study in the Journal of Human Evolution produced a digital reconstruction. Rupe reports that its author told him by e-mail in February 2025 that he had built it without reference to the earlier reconstructions, and then found it resembled Schmid and Häusler's. Crucially, that team reconstructed the pelvis without restoring the damaged joint at all — leaving the visible gap — and still got a human-shaped pelvis. Their dataset is public.⁴⁶
4. Casts of the hip bone with the unrestored blade can be bought online. The claim is checkable by anyone with a hundred dollars.

A creationist with an earth-sciences PhD, writing in the flagship creationist journal, is telling other creationists to stop using this clip.  That is the most useful fact in this part, because it means the argument can be retired without conceding anything to the other side.

What survives of the criticism, stated precisely

Three things, and only these three.

1. A reconstruction that converts an ape-like pelvic flare into a human-like one is the primary evidence for the headline locomotor claim. That is a real dependency and worth stating.
2. Named anatomists have twice proposed to redo the reconstruction differently — Häusler and Schmid, and Berge. None of them is a creationist, and none concluded "ape."
3. A separate claim, that the procedure was shown on television before it was documented in the technical literature, is not printed here as fact. Checking it requires opening three specific papers from 1979, 1982 and 1986, and nobody on this project has yet done so.³¹

The rule this produces

The honest version of this argument is not "they cut up the fossil."  It is:  when a reconstruction changes the answer, it has to be documented well enough for somebody else to redo it and disagree.

Lovejoy's was documented on film.  The 2018 team's was documented as a public 3-D dataset.  The second standard is the one to praise — and it is the same standard that made the Brasilodon correction in Part 3 possible.

Part 9 — Our own pictures, audited

An article about misleading pictures that publishes misleading pictures is finished before it starts.  So this project's own fossil image library was inspected for this article.  Two findings, both first-hand, both against us.

zachelmie-prints.jpg — do not publish

The file is a 128 KB image, saved in May 2015, held in this project's fossil-record image folder.  Opened and read directly for this article, it turns out to be a screen capture of a page from nature.com:  the masthead line "Published online 6 January 2010 | Nature | doi:10.1038/news.2010.1," Rex Dalton's byline under the headline "Discovery pushes back date of first four-legged animal," a photograph of the Polish trackway slab, and a yellow highlight added by whoever saved it.

Three separate reasons it must never appear on this site.

1. It is a copyright problem. It is a screenshot of an in-copyright news page including its photograph. Under this project's own rules it is excerpt-only and cannot be republished.
2. The date on it is superseded. The capture says about 395 million years. The Zachełmie age has been published three times by the same research group — 397, then 395, then about 390 million years — so the image freezes the middle value of three.¹²
3. The highlighted sentence is superseded too. The highlight sits on the words "more than 18 million years before tetrapods were thought to have evolved." That is the 2010 figure. The same group now uses about 14 million years, and creationist writing is still recycling the 2010 number sixteen years later.¹²

The yellow highlight — added by us — is sitting on the one number in the image that is out of date.  There is no way to publish that image honestly, and a replacement figure built from CC-BY sources already exists for the companion article.

limb-evolution.jpg — usable only with a caption that says what it is

Same folder, 251 KB, same 2015 date.  It is a fin-to-limb diagram running from a ray-finned fish through Eusthenopteron, Panderichthys, Acanthostega, Ichthyostega, Tulerpeton and Greererpeton to modern tetrapods, with arm bones colour-coded and digit counts labelled.

It mixes three kinds of image on one line without telling the reader which is which.  Actual photographs of fossil slabs.  Black-and-white skeletal line drawings.  And full-colour flesh-and-skin life restorations of Panderichthys, Tulerpeton and Ichthyostega.

The visual grammar puts a photograph of a rock and a painted animal at the same evidential level.  That is the exact failure this article is written to name, and it is in our own folder.  Its provenance is also unestablished.

There are two options:  do not use it, or use it as the example, captioned so the reader is told which panels are photographs of bone and which are somebody's painting.  The second is better and cheaper.

The standing rule this produces

Every image published in this series carries a caption saying, in plain words, which parts are bone.

If the answer is "all of it," say so — that is a point in the picture's favour.  If part of it is inferred, say which part.  This costs one sentence per figure, and it removes the strongest tu quoque a hostile reader has.

Objections and responses

The objections below come from both directions, and the ones from our own side are answered as sharply as the others.

1. "Every fraud you list was caught by scientists. That is an argument for the system."

Correct, and this article says so in its first paragraph.  It is the reason the fraud argument is not made here.  Piltdown was caught by a chemical test its exposers designed.  Nebraska Man was retracted by its own co-describer.  Archaeoraptor was scanned before the magazine printed it.  Xenodens was accused of forgery and cleared by CT within a year.

The argument this article does make is different, and the fraud cases do not carry it.  It is about how long an honest error can live, how much of a picture is inference, and what a purchased specimen can prove.

2. "So one Nature figure had bad claws. Errors get fixed. What is the complaint?"

The complaint is the timing, and the reason for it.

Casineria's correction took twenty years, and it required somebody to fly to Edinburgh and spend a day with the only specimen.  Oculudentavis's took four months, and it required somebody to download the original authors' own CT files.  Same field, same decade, same kind of error.

Correction speed tracks data availability, not integrity.  That is a testable claim about a process, and the practical conclusion — publish the scan data — is one a skeptical reader will already agree with.

3. "The 81-piece Confuciusornis slab is a preparation fact, not an evidence problem."

Half right, and the half that is right is why this article prints it.  The 81 pieces are not a scandal; they are what a thinly bedded fossil looks like after it is dug and repaired.

The evidence problem is that the surface of a prepared slab does not show you which pieces are which.  That is exactly what the CT scans were for, and it is why the same technique cleared Xenodens and convicted Archaeoraptor.  The complaint is not that fossils are mosaicked.  It is that the mosaic is invisible until somebody scans it, and most specimens have not been scanned.

4. "You are cherry-picking corrections. Most fossils are described correctly."

Almost certainly true, and this article cannot show otherwise.  Nobody has computed an error rate for fossil descriptions, the literature has no such number, and this article does not invent one.  What it can show is that errors occur at the top of the field, on the taxa carrying the most argumentative weight, and can survive for decades.  That is a statement about the tail of the distribution, not the middle — and the tail is where the famous fossils live, because a taxon becomes famous by being extreme.

5. "Creationists have used fraud arguments dishonestly for a century. Why trust this?"

Because this article kills three of them by name, in our own voice, before making any argument of its own.  The Lucy power-saw clip is refuted by the narration in the same clip.  Nebraska Man was not at the Scopes trial, and that was checked in two independent copies of the trial transcript.  Piltdown's exposers said its removal clarified human evolution, and their sentence is printed here in full.

And one quotation that would have helped is deliberately omitted because its only available copies run through advocacy websites.

6. "The commercial-trade argument is an ethics complaint dressed up as an epistemology complaint."

The ethics are real, and this article reports them.  But the epistemic point stands on its own and does not need them.

A dealer who sorts bones by matching size and shape before selling them manufactures the pattern that a scientist then reads as evidence of association.  That is not an accusation against anyone.  It is a statement about what a matching set of purchased bones can prove, and mainstream specialists printed it themselves.

7. "You cite Sharpe et al. on Xenodens forgery, then say they were wrong. Which is it?"

Both, and the sequence is the point.  A forgery allegation was published in a respectable journal, with a specific proposed test.  The test was run.  The allegation failed — the supposed adhesive was a standard conservation resin that every prepared fossil carries.

A forgery allegation is a claim that has to survive testing too.  Anyone who cites the accusation without the rebuttal is doing what this article accuses others of doing.

8. "If reconstructions are this soft, no fossil argument is worth anything — including yours."

This is the objection this article most wants to head off, and it is wrong.

The 2025 dinosaur study makes the distinction cleanly.  Absolute masses vary threefold across investigators; the rank ordering of 52 taxa by mass holds across methods at a correlation above 0.99.  Soft absolute claim, robust relative claim.

Fossils reliably support claims about presence, order and comparison:  this animal existed, it occurs below that one, it is larger than the other.  They support claims about soft tissue, colour, behaviour and precise ancestry far less reliably.  A reader who takes this article as licence to ignore fossils has misread it.

9. "Creationist arguments rest on single specimens too."

They do, and the same test applies.  A creationist argument built on one anomalous fossil, one photograph, or one screenshot is worth exactly what Casineria was worth in 1999.  Part 9 applies that rule to this project's own image folder, and both files audited there failed it.

What this licenses, and what it does not

The failure mode of a chapter like this is obvious and fatal:  a reader concludes "fossils prove nothing."  That is false, unearned, and would embarrass this site the first time it needed a fossil for its own argument.

It does not license

• "The fossil record is fake." Every object in this article is real. A real orang-utan, real medieval humans, a real bird, a real dinosaur, a real peccary, a real Carboniferous tetrapod.
• "Scientists are dishonest." Two deliberate forgeries appear here, and in both the forger was outside science: a Sussex antiquarian and an unidentified Chinese preparator. Every other case is honest error, and every one was caught by mainstream scientists — usually before any creationist noticed.
• Using Piltdown as a live argument. Its exposers wrote that removing it "clarifies very considerably the problem of human evolution."
• The Nebraska Man story as usually told. Not at Scopes, not called an ape-man by its describer, retracted in Science by its own co-describer.
• The Lucy power-saw clip. The narration says it was a plaster copy, and a creationist with a geology PhD says in print that the fraud allegation does not stand up.
• Treating a correction as a scandal. Brasilodon, Tikitherium, Oculudentavis and Casineria are the system working. Slowly, in one case. But working.

It does license

One question, asked of any reconstruction:  which parts are bone?

That is the whole article in a sentence.  It is not skepticism.  It is a request for a caption, and the professional version of it was written by a mainstream paleontologist in an open-access journal:  a restored model "should therefore be regarded as a working hypothesis, which can be amended with the emergence of new data."²

Five supporting claims follow from it, each traceable to a source above.

1. A published figure can be wrong, and the error can outlive the paper by decades. Casineria's claws were drawn wrong for twenty years. Bakker's foreshortened Barosaurus tail became a short tail in six later publications over twenty-seven years.
2. How much of a picture is fixed by bone is quantifiable, and the answer is often "not much." Six to eighteen tonnes, on the same Tyrannosaurus skeleton.
3. A famous taxon is often thin, and the ones carrying the most weight are the thinnest. One headless immature specimen. One tooth. One maxilla bought from a farmer. One lost vertebra known from an 1878 engraving.
4. Where the object came from is part of the evidence. A specimen bought from a dealer arrives pre-assembled by someone paid to make it look complete.
5. Correction speed tracks data availability, not integrity. Four months when the scan data were public. Twenty years when checking required a plane ticket.

The last one is the strongest practical conclusion in this article, and it is the one a skeptical reader will agree with fastest.  The argument for open data is an argument this site and its critics can make together.

What would change this verdict

• The full text of Paton, Smithson and Clack 1999, so figure 3b can be examined directly rather than through Marjanović and Laurin's description of it. As it stands the figure error rests on one side of a two-party dispute.
• A published reply from Paton, Smithson or Clack. None could be found. Absence of a reply is not agreement, in either direction.
• A reconciliation of the two published accounts of the Archaeoraptor slab's composition — one individual of Yanornis, or two to five specimens. Nobody has done it.
• A reply from Sharpe, Powers and Caldwell to the 2025 CT rebuttal on Xenodens. The rebuttal is nine months old, and this should be re-checked before publication.
• A measured error rate for fossil descriptions, which does not exist in the literature. Without one, this article can describe the tail of the distribution and not its middle, and it says so.
• A copy of Storrs Olson's 1999 letter from a primary or archival source, which would let the most-quoted contemporary criticism of the Archaeoraptor affair be used at all.
• All Yesterdays (2012), so its reverse experiment can be reported rather than mentioned.
• Established provenance for limb-evolution.jpg, without which it cannot be published even as this article's own example.

Sources

Research dossier, with every quotation, its verdict, the live database queries and the sources that could not be reached:  research/reconstruction-and-fraud.md.

What was read in full, and what was not.  Read as full open-access texts:  Rowe et al. 2016, Marjanović and Laurin 2019, Rawson et al. 2024, Longrich et al. 2025, Dunne et al. 2022, Evers et al. 2015, Dempsey et al. 2025, Lautenschlager 2016, Witton, Naish and Conway 2014, Carpenter 2018, De Groote et al. 2016, Weiner, Oakley and Le Gros Clark 1953, Rupe 2026, and the Osborn and Gregory Nebraska Man papers.  Abstract only:  Paton et al. 1999, Rowe et al. 2001, Averianov and Voyta 2024, Sharpe et al. 2025, Xu et al. 2000, Zhou et al. 2002, Bolet et al. 2021, Sereno et al. 2022, Konietzko-Meier et al. 2020.  Every footnote below says which.

Two sources that would have strengthened the article could not be obtained at all:  the 1922 New York Times report of the American Museum's reaction to the Illustrated London News drawing, and All Yesterdays.  Both are named where they are used.