The Reptile Jaw and the Mammal Ear

"Creationists therefore need to accept this fact.  It certainly CANNOT be said that traditional creation theory expected (predicted) any of these fossil finds." — Kurt Wise, 1995, a young-earth creationist with a Harvard paleontology PhD, writing in a creationist journal.¹

Carl Reichert worked out where the mammal's ear bones came from in 1837 — twenty-two years before Origin of Species, and before anyone had found a single mammal-like reptile.  He did it by looking at embryos.²

Summary

You have three tiny bones in each middle ear.  A reptile has one.  The other two are in the reptile's jaw.

The claim is that those two bones migrated — that over a long series of animals, the bones forming the back of the jaw joint shrank, detached, and became the hammer and anvil of the mammalian ear.

This article concludes that the claim is well supported, and that no young-earth objection to it survives examination.

That is not the conclusion this site would prefer.  It is the conclusion the evidence forces, and saying so is the point.  A series of articles in which every transitional sequence conveniently fails would not be worth reading.  This one is the strongest case the other side has, so it goes first.

The people who work on it say the same thing in the same words.  The 2025 PNAS paper that modelled the intermediate ear opens by calling this transformation "the most thoroughly documented morphological transformation in mammalian evolutionary history."³  That is a strong claim from an interested party, and the rest of this article is an attempt to check it rather than to repeat it.

Why this sequence is different from the others

Four things separate this case from the horse series, the whale series, or any other line-up of fossils.

The prediction came before the fossils.  Carl Reichert proposed the homology in 1837, from embryos, before any mammal-like reptile had been described and twenty-two years before Darwin published.²  Ernst Gaupp extended it in 1913 in a 416-page monograph.⁴²  The fossils began arriving in 1912 and have kept fitting for over a century.²

Since 1990 the same homology has been confirmed a third time, in a data domain neither Reichert nor Darwin could have imagined.  The bones share a single cartilaginous condensation that later subdivides.⁵  They share a pharyngeal-arch origin traceable by neural-crest fate mapping.⁵  And the same gene — Bapx1 — marks the jaw joint in a chick and a python and the malleus–incus joint in a mouse.⁵

The intermediate stage works, and it is not hypothetical.  For part of this sequence, animals had two jaw joints at once — the old one and the new one, side by side, both articulating.  Diarthrognathus is named for exactly that, though its family reached the double joint separately from the line to mammals.⁶⁷  Morganucodon, an early mammaliaform, has it, and the paper describing the load transfer says so in those words.⁸

And a newborn marsupial still does it.  For weeks after birth it hinges its jaw on the malleus–incus joint, before the adult mammalian joint takes over.⁹

At the one stage that has been modelled, the animal could hear.  A 2025 finite-element study of Thrinaxodon found that its ear-in-the-jaw arrangement already worked as an airborne, eardrum-based listener, with best sensitivity near 1,000 Hz — while pure bone conduction did not reach the audibility threshold at all.³

The material is not thin.  The iDigBio specimen aggregator returns 7,636 catalogued records for Morganucodon alone, 7,460 of them at the Natural History Museum in London, nearly all from Welsh fissure fills.¹⁰  That is three times as many specimens as Dimetrodon.  "A handful of bones arranged by an evolutionist" is not a weak argument here.  It is a factual error, and Part 2 prints the counts.

This is a class-level transition, and that matters

A creationist reading this should know precisely what rank is at stake, because the answer decides whether the sequence proves anything.

Young-earth biology generally accepts common ancestry within a created kind — usually somewhere around the family level.  A series of horses is not a problem for that view.  A series of dogs is not a problem.

This one is.  The Paleobiology Database's own Linnaean rank codes put the start of the sequence in the order Pelycosauria and the end in the class Mammalia, with 809 named non-mammalian synapsid genera in between, 225 of them cynodonts.¹¹¹²

That also answers a fair objection to the way Stephen Jay Gould is usually quoted.  Gould wrote that transitional forms are "generally lacking at the species level, but they are abundant between larger groups" — and "larger groups" is vague enough that a creationist can reasonably ask whether he meant anything above a family.¹³  In that same essay, on the following page, his worked example is this sequence.¹³  So in Gould's case "larger groups" means class.

One correction runs with that, and our own side should make it first.  Synapsids were never reptiles.  "Mammal-like reptile" is a legacy of nineteenth-century classification and is deprecated in the technical literature.  The synapsid line separated from the sauropsid line back in the Carboniferous, so Dimetrodon is not a reptile that became mammal-like — it is an early member of our own branch.  The rank claim does not depend on that correction, because it never depended on the starting group's name.  It depends on the ending group's rank, and Mammalia is a class under every scheme, traditional or cladistic.¹¹

What this article concedes anyway

The genuine weaknesses, stated before the case is made.

• Almost none of these animals is an ancestor. The sequence is a reconstruction from a tree, not a line of parents and children. Most of the named taxa are side branches that retained an intermediate condition.
• A 2024 paper overturned the published account of the taxon closest to mammals. Brasilodon was reported for about twenty years to have an incipient mammalian jaw joint. Micro-CT scanning showed it has neither the dentary condyle nor the squamosal glenoid, and articulates solely by the old joint.⁷
• The new jaw joint evolved more than once, and the ear detached from the jaw at least twice and possibly up to seven times.¹⁴⁷ This is a genuine complication and mainstream paleontologists debate it in the open. Part 5 gives it in full.
• The eardrum has never been fossilised. It is soft tissue. Its position is inferred from the shape of the angular's reflected lamina, and the 2025 model that "hears" is modelling a tympanum nobody has ever seen.³
• The trend is not as smooth as it is usually drawn, and a 2023 paper measured that. Harano and Asahara tested the classic claim that the dentary steadily enlarged at the expense of the postdentary bones across all non-mammalian synapsids. Relative dentary area does show a significant trend. Relative dentary length — the measure that actually tracks postdentary reduction — does not, and the reconstructed history is not unidirectional.¹⁵ Part 1 gives the numbers and what they do and do not overturn.
• The textbook adaptive story was falsified in 2023 — by evolutionists. The standard account is that consolidating the jaw into one bone made the skull stronger. Modelling found no increase in cranial strength; stresses rose in the zygomatic region and mechanical advantage fell.¹⁶ That also kills an aside Kurt Wise made in 1995, since he repeated the same adaptive story.¹
• The order in the rock fits the tree well, not perfectly. Read in standard tree order, eight early synapsid groups show one small reversal, and a random order would do as well about once in 670. Part 1 gives the test.
• A young-earth observation has been borne out. Kurt Wise pointed out in 1995 and 2006 that several separate families grow more mammal-like in parallel, and a 2024 Nature paper confirms it for the jaw joint.¹⁷⁷ It does not rescue his model, which still has no map. Part 6 gives both halves.
• Three mainstream dating and anatomy claims in this literature are simply wrong, and this article names them in Part 8. One is still live in the Paleobiology Database.

And one argument this article retires from our own side

The waiting-time argument does not apply here, and it should not be used.

The objection says population genetics cannot deliver the required changes in the time available.  The best-developed version is Sanford and colleagues' 2015 simulation, which found that establishing a specific two-nucleotide string in a hominin-sized population took on average 84 million years, and a specific five-nucleotide string about two billion.¹⁸  That objection is serious in other contexts and this site treats it seriously elsewhere.¹⁹

It is not usable in this one, for two specific reasons.

First, the model needs a target and there is no target here.  Sanford's simulation waits for a pre-specified string of nucleotides.  What this sequence involves is quantitative traits — bone proportions, positions, sizes.  "The quadrate gets smaller and shifts medially" is not a string.  Many different mutations at many different loci move a trait like that in the same direction, so a model built to wait for one exact sequence is the wrong model.

Second, and more decisively: nobody has ever counted the coordinated genetic changes this transformation required.  There is no published estimate.  Without that number, any waiting time produced for this case would be manufactured.  Saying so plainly is worth more than a calculation nobody can defend.

For scale, and cited as the mainstream figures under discussion rather than as evidence: the interval from the oldest ophiacodontid to the oldest secure morganucodont is about 109 million years, and these were small animals whose living equivalents breed inside a year.²⁰  Sanford's 84-million-year result consumed roughly 4.2 million generations; the synapsid window offers something on the order of 10⁸.  That comparison rests on an assumed generation time nobody has published for these animals.  It is a scoping exercise, not a result, and this article will not dress it up as one.

Radiometric and absolute dating are not used as evidence anywhere in this article.  That is a constraint across this series.  Every argument here works from relative order — what sits above what — which young-earth geology accepts.  Where an absolute age appears, it is the mainstream figure under discussion, not something a reader is being asked to believe.

Part 1 — The bones, and where they went

A reptile's lower jaw contains several bones.  A mammal's contains one — the dentary.  A reptile hears with one middle-ear bone, the stapes.  A mammal has three: the stapes plus the malleus and incus.

The proposal is simple and specific.

The stapes is not part of the story.  It was already an ear bone before any of this began and it stays one.  What is new in mammals is that the jaw joint itself became the first two links of the ossicular chain, and a wholly new jaw joint was built out of two membrane bones to replace it.

That last point defeats a common framing of the problem.  The mammalian ear was not made by converting a one-ossicle reptile ear into a three-ossicle mammal ear, because the common ancestor did not have a one-ossicle ear to convert.  The stem amniote had no tympanic middle ear at all.  Its stapes was a stout strut bracing the skull during biting, with no evidence of an airborne-sound role.  Eardrums arose separately in the different amniote groups — and again in frogs.²¹  Mammals built a three-ossicle tympanic ear from the start.

The three named stages, and which fossil is at each

The modern terminology comes from Meng and Mao's 2024 review, written by the authors of the two Nature papers it summarises.⁸  Three stages, each with a fossil that shows it:

1. Mandibular middle ear. The postdentary bones still sit in a trough on the inner face of the dentary. They serve hearing and the jaw joint at the same time. Morganucodon, Dianoconodon, Feredocodon.
2. Transitional mammalian middle ear. The postdentary unit has detached from the dentary and works as an ear, but the malleus is still anchored to an ossified Meckel's cartilage lying in the Meckelian groove of the dentary. Chewing and hearing are still mechanically coupled. Liaoconodon.
3. Definitive mammalian middle ear. The ossified Meckel's cartilage is gone and the ossicles are free. Origolestes is described as catching the moment of separation. Every living adult mammal has it.

Read as a list of changes, the transformation runs like this, and every step is attested in some fossil:

• The dentary enlarges until it forms nearly the whole lower jaw, growing a coronoid process, an angular process, and finally a condyle.
• The postdentary bones shrink — articular, prearticular, angular, surangular — but keep their relative arrangement, which is why they stay recognisable.
• A dentary–squamosal contact appears alongside the still-functioning quadrate–articular joint.
• Both joints work at once. Part 4 is about why that step is the one that cannot be explained away.
• The postdentary unit detaches, leaving first a postdentary trough on the dentary and then only a Meckelian groove.
• The ossified Meckel's cartilage persists as the last mechanical link, then disappears.

The step-by-step evidence, not just the endpoints

The transfer of load from the old joint to the new one is visible bone by bone, and the geometry is the evidence rather than the assertion.

In Dianoconodon, the postdentary bones, quadrate and quadratojugal are all further reduced, the stapes has slim crura, and the quadrate has shifted medially with its articular trochlea set on a diagonal.  Meng and Mao's point is mechanical: a trochlea in that orientation could not have carried much chewing stress.⁸

In Feredocodon, the quadrate has lost its trochlea altogether and the quadratojugal is gone.  The old joint, in their words, "has completely lost its load-bearing function and works exclusively for hearing."⁸  They identify the key mechanical change across the whole series as the medial shift of the reduced quadrate relative to the articular.

Liaoconodon preserves the stage after detachment but before independence.  Its ectotympanic and malleus "have lost their direct contact with the dentary bone but still connect the ossified Meckel's cartilage."²²  That is a bony strut running back along the line where the jaw used to hold them.  The same ossified element is known in Yanoconodon, in gobiconodontids, and in the spalacotheroid Maotherium.⁸

That is not two endpoints with a gap in the middle.  It is bones in the jaw doing both jobs, then a quadrate reoriented so it can only do one, then a detached ear still tethered by bone, then a free ossicular chain.

What is not preserved, and is therefore inferred

Stated here rather than buried, because every one of these is a place a skeptic can push.

• The eardrum is soft tissue and has never been fossilised in any synapsid. Its presence, size and attachment are inferred from the reflected lamina of the angular and from the retroarticular process. The 2025 finite-element study models a tympanum; it does not observe one.³
• Meckel's cartilage is only visible when it ossified, which is the exception. Where it stayed cartilage, its presence is inferred from a groove.
• Absence of a postdentary trough does not by itself prove the ossicles had detached.⁸ Yet trough-based scoring drives a large share of the phylogenetic signal in taxa with no preserved ossicles, which makes those placements weaker than they look.
• Whether the mammalian and non-mammalian eardrums are homologous is still disputed, on the position of muscles relative to the Eustachian tube.⁵ This article does not take a side.
• Hearing sensitivity, diet, nocturnality and behaviour are all reconstructions.

The order in the rock matches the order in the tree

This is checkable from relative position alone, and it was checked for this article rather than quoted.  First appearances were computed from the Paleobiology Database, using only occurrences whose age bracket resolves to 15 million years or better.²⁰

Read down the cladistic order from Ophiacodontidae to Cynodontia and the first appearances fall from 314.60 to 264.28 million years ago, with one small reversal.  The rows are in the order of the standard tree, in which gorgonopsians are the sister group of therocephalians plus cynodonts, so anomodonts branch off first.²³²³  Read that way, the anomodonts' first record (274.40) comes 2.90 million years after the gorgonopsians' (277.30) on the mainstream scale.  That gap is smaller than the 15-million-year age brackets the test admits, and the gorgonopsian figure rests on two Russian records with no formation recorded, so it is noise rather than a result.  Listing gorgonopsians first would hide that reversal, which is why the order of the rows has to be stated.

Only the first eight rows are a test.  Epicynodontia, Eucynodontia and Probainognathia sit inside Cynodontia, so none of them can have an older first record than Cynodontia itself.  Their falling dates were guaranteed and prove nothing.  The eight informative clades give five steps down, one tie and one step up.  Shuffle those eight first appearances at random and an order this good or better turns up 60 times in 40,320, about once in 670.²⁴  That is a strong fit, and it is the honest size of it.

That is not a claim about absolute dates.  It is a claim about sequence, which holds from superposition regardless of what anyone thinks of decay constants.  No hydrodynamic or ecological sorting model has been shown to predict that particular order.²⁵

The exceptions, printed rather than smoothed.

• Tritylodontidae (205.70 Ma) appears before Tritheledontidae (201.40 Ma), whereas most trees place tritheledontids closer to mammals. Both are side branches, so the order between them is only weakly predicted — but it is an inversion and it is ours to report.
• Mammaliaformes is left out of the table above on purpose. Its oldest record in the database is 237 Ma, from the Tiki Formation of India, and that record is almost certainly bad. It is Gondwanadon tapani, a single tooth from the same formation and collection as Tikitherium copei — and Tikitherium was redescribed in 2024 as a Neogene shrew. See Part 8.²⁶
• Brasilodontidae's oldest database record is 251.9 Ma, from the Panchet Formation of India: Panchetocynodon damodarensis, a fragmentary left mandible with five damaged postcanines, about 25 million years older than the Brazilian brasilodontids. Whether that family assignment is sound could not be established for this article, so it is used as evidence by nobody here, in either direction.²⁷
• Biarmosuchia, Dinocephalia and Gorgonopsia all tie at 277.30 Ma on Russian occurrences with no formation recorded. A tie is not an inversion, but the resolution there is poor and the tie should not be read as a result. The gorgonopsian figure is the one that produces the reversal above.

There is a further, subtler check.  Sidor and Hopson measured how much anatomical change each lineage shows against the length of its ghost lineage — the stretch of time a lineage must have existed, because its sister group is already present, but from which no fossils are known.  They found a statistically significant positive relationship, and, in the sentence that carries the most weight, "in no case was a long ghost lineage associated with few apomorphies."  Their conclusion was that "gradual character evolution predominated."²⁸

That study cuts both ways and this article prints both.  The pro-sequence reading is that morphological change tracks elapsed time: big anatomical jumps do not fall out of short intervals, which is what gradual evolution predicts and is not what a sorting artefact would produce.  The anti-sequence reading is that the study's own unit of analysis is missing time.  The correlation is between counts of derived features and the length of intervals from which nobody has any fossils.  A young-earth reader can fairly say the smoothness is partly a property of how the gaps were filled in by inference.

The measured result that qualifies the trend, and it is recent

The claim everybody repeats — Kurt Wise repeats it too — is that the dentary steadily enlarged while the postdentary bones steadily shrank, all the way up the synapsid record.  In 2023 Tomohiro Harano and Masakazu Asahara tested that with phylogenetic comparative methods, which nobody had previously done.¹⁵

Their method: measure the dentary and the whole lower jaw in side view from published illustrations of 53 non-mammalian synapsid taxa, 41 of them therapsids; map the measurements onto a time-calibrated supertree; and fit eight models of trait evolution, including a Brownian-motion model with a directional trend parameter.  Areas were obtainable for 43 taxa.  Two length measures were taken as well, because area can grow without the postdentary bones shrinking at all — a taller dentary does it.

Their results split.

Read that carefully, because it is the most important qualification in this article.  Relative dentary area really does increase across non-mammalian synapsids, and the directional model beats every alternative.  But the length measures — the ones that track the dentary growing at the expense of the postdentary region — show no significant trend.  The authors' own inference is that the area trend is driven by the dentary expanding vertically, growing a coronoid process and a taller front end for muscle attachment, and that this is about bite force rather than about the ear.  Their ancestral reconstructions show relative dentary length shorter than the immediate ancestor's in a number of lineages, so the change was not unidirectional.¹⁵  Their conclusion, in the abstract, is that they found no evidence of a trend toward dentary enlargement at the expense of the postdentary bones across non-mammalian synapsids as a whole.¹⁵

What that does and does not do.

It does not overturn the transformation.  Every fossil described earlier in this Part still has the anatomy it has, and the dentary really does end up as the only bone in the jaw.

What it does is relocate the change.  Harano and Asahara's own reading is that the postdentary reduction was not a slow 150-million-year drift shared by all synapsids.  It was concentrated in the transition from non-mammalian cynodonts to early mammals, driven by selection that operated in that lineage and not in the others.  Their length-2 measure was best fitted by a model whose parameter value points toward change concentrated at branching events followed by long stasis, which is a punctuational rather than a gradual signature — though they caution that their tree contains a limited number of taxa and the branching events are unevenly distributed.¹⁵

It also partly vindicates a creationist objection, and this article says so.  John Woodmorappe argued in 2001 that the mammalian characters do not climb smoothly up the sequence.  On this particular measurement, over the whole of Synapsida, he is closer to right than the textbook picture is.  Part 7 explains why his argument still fails — but it fails for reasons other than this one, and he should get the credit that is his.

Part 2 — How much material there actually is

The commonest creationist claim about this sequence is that it rests on scraps arranged by an evolutionist.  That claim is testable, and two open databases test it.  Neither is run by anyone with a stake in this argument.  iDigBio aggregates digitised museum collection records.  The Paleobiology Database records fossil occurrences by collection and stratigraphic interval.  The counts below were pulled directly from both.¹⁰²⁰

Read the two columns differently, and do not add them.  A database occurrence is one taxon in one collection.  An iDigBio record is one catalogued museum object — usually one bone, tooth or jaw.  Neither is a headcount of animals.

And iDigBio's bias has to travel with its numbers.  Its holdings are overwhelmingly North American and British; Chinese, Russian, South African, Brazilian and Argentine collections are largely absent.  A zero means "not digitised into iDigBio," never "no specimens exist" — Hadrocodium, Yanoconodon, Liaoconodon and Brasilodon all return zero and all of them exist.  The counts are used here only where they are large and positive, because there they are a lower bound and therefore safe.¹⁰

The counts

A dash means the genus was not separately queried in iDigBio, not that it returned zero.

A correction to an earlier version of this table.  The occurrence column was first built by summing the database's n_occs field across the genus row and its species rows of a taxa/list query, which counts most records twice — Morganucodon was printed as 24 and Riograndia as 8.  The figures above are occurrence rowcounts from occs/list, queried directly, and they are the ones that agree with the counts printed elsewhere on this site.

The finding that matters most

Morganucodon is not a scrap.  iDigBio holds 7,636 catalogued specimen records for the genus.  7,460 are at the Natural History Museum, London, and 7,459 of those are from the United Kingdom.

Sampling the records shows what they are.  Individually numbered specimens — NHMUK PV M 109031 through PV M 109035, for instance — all labelled Morganucodon watsoni, all from Pontalun 3 fissure in Pontalun Quarry at Ewenny, south of Bridgend, in South Wales.  These are Early Jurassic fissure fills in Carboniferous limestone.  A GBIF query restricted to fossil specimens returns the identical figure, 7,636.¹⁰

Kuehneotherium, from the same Welsh fissure system, returns 1,465 in iDigBio and 1,307 in GBIF.

So the two mammaliaforms sitting at the hinge of this transition are among the better-sampled fossil vertebrates in any museum.  The material is public, numbered, and searchable by anyone with a browser.  Any article on either side that describes the earliest mammals as known from "a few teeth" is refuted by a database query that takes ten seconds.  For comparison, Dimetrodon — the animal most people can name — returns 2,522, a third of Morganucodon's total.

Where the material is genuinely thin, stated in the same breath

The same counts show where this sequence is weak, and the article that prints the 7,636 has to print these too.

• Hadrocodium wui is one skull. Holotype IVPP 8275. Skull length 12 mm, estimated body mass about 2 grams.²⁹ One database occurrence, zero iDigBio records. A single specimen carrying a large share of the argument about when the ear detached is a real weakness, not a rhetorical one.
• Diarthrognathus broomi — the animal whose name means "double jaw joint," and the textbook poster child for the intermediate stage — has one database occurrence and two iDigBio records, one of which is the genus-level entry.⁶
• Teinolophos trusleri, the Early Cretaceous stem monotreme at the centre of the independent-origins dispute in Part 5, has one database occurrence and four iDigBio records. The argument about it turns on the interpretation of a groove in a partial dentary.³⁰
• Brasilodon and Riograndia, the two taxa a 2024 Nature paper used to overturn the previous account of the oldest dentary–squamosal contact, have one and four occurrences. That paper worked from 25 micro-CT-scanned specimens in total — nine Brasilodon, nine Riograndia, seven Oligokyphus — scanned at voxel sizes of 0.0057 to 0.026 mm over roughly 1,500 to 3,000 slices each, mostly loaned from the Universidade Federal do Rio Grande do Sul.⁷ That is a serious modern dataset built on a small number of individuals, and both facts are true at once.

One thing this article deliberately does not print

A specimen table would normally carry a holotype column.  This one does not, and the reason is worth stating because it is the kind of thing that gets copied forward once somebody guesses.

Holotype catalogue numbers were verified from a primary source for only three items in this whole sequence: Hadrocodium wui (IVPP 8275), and the Brazilian Brasilodon and Riograndia material in the 2024 scanning study (catalogue numbers of the form UFRGS-PV-596-T).²⁹⁷  The Paleobiology Database does not carry type-specimen numbers at all, and the remaining original descriptions were not opened.

One near-miss shows why the caution is needed.  The Thrinaxodon specimen in the 2025 hearing study is BP/1/7199, with an articular grafted in digitally from UCMP V40466.³  That is the study specimen.  It is not necessarily Seeley's 1894 holotype, and this article does not claim it is.

No holotype number appears anywhere in this article that was not read out of the paper that assigned it.

Part 3 — The prediction came before the fossils

This is the strongest single part of the case and it is the part most often left out.

The claim that the mammal's ear bones are the reptile's jaw bones was not read off the fossils.  It was worked out from embryos.  It was worked out before Darwin published, before the mammal-like reptiles were known as a body of evidence, and by men who were not evolutionists.

The two names

Carl Bogislaus Reichert, 1837.  Writing in the Archiv für Anatomie, Physiologie und wissenschaftliche Medicin, Reichert traced the pharyngeal — then called visceral — arches through mammal and bird embryos, and concluded that the malleus and incus are the same structures as the articular and quadrate of other vertebrates.³¹

Ernst Gaupp, 1913.  A 416-page monograph on what he called die Reichertsche Theorie, which added the distinction the whole subject now runs on: a primary jaw joint, malleus to incus, which is the articular-to-quadrate joint of everything else, and a secondary jaw joint, dentary to squamosal, which is unique to mammals.⁴

The composite is called the Reichert–Gaupp theory in the modern literature.²  A peer-reviewed review states the priority plainly: "Reichert (1837), and later Gaupp (1913), were the first to recognise the homology of the reptilian jaw joint and the mammalian middle ear elements based on embryological studies."²

A citation trap worth flagging, because it is in a widely cited paper.  Anthwal, Joshi and Tucker's 2013 review prints "Reichert, 1937" in the text and a reference-list entry reading "1937;1837:120–220," and dates Gaupp to 1912.  The correct dates are 1837 and 1913.⁵

Why the order matters

Reichert published twenty-two years before On the Origin of Species.  He was not testing evolution.  He was an idealist morphologist comparing developmental sequences, working in a tradition that had no theory of descent to defend.

At that date the fossils did not exist as evidence.  Owen's first South African therapsids were described in the 1840s.  The cynodont material that actually shows the transformation was collected mostly from the 1890s onward.  The fossil work that tested Reichert's idea came later still: Watson on the jaw musculature of Cynognathus in 1912, Adams in 1918, and then the postwar wave that produced the real intermediates — Diarthrognathus in 1958, the Morganucodon work from 1963, Probainognathus in 1969 and 1970.²  Broom's 1912 paper on Dicynodon is titled, in part, "on the homology of the mammalian auditory ossicles" — a paleontologist explicitly checking an embryologist's claim against bones.

So the order is: embryology in 1837, then fossils that fit it from 1912 to 1970, then a further half century of fossils that keep fitting.  That is a hypothesis surviving a test in a data domain it was not built from, which is the only kind of test that carries much weight.  A young-earth reader who has been told this sequence was assembled by people who already believed in evolution should sit with that date.

The modern developmental test, which is a third confirmation

Everything above is nineteenth- and early twentieth-century.  Since 1990 the homology has been checked again with tools that did not exist when the fossils were found, and it has passed three separate times.⁵

Shared condensation.  The malleus and incus, like the articular and quadrate, begin life as a single cartilaginous condensation that then subdivides.  The stapes forms from a separate condensation.  So the two bones that the fossil argument says were once one joint are, in a living embryo, still one block of cartilage before they separate.⁵

Shared arch origin.  Neural-crest fate mapping — quail-into-chick chimaeras, and reporter mice that permanently label cells by which pharyngeal arch they came from — puts the malleus and incus in the first arch and the stapes in the second, exactly as the homology requires.⁵

Shared joint gene.  Bapx1 marks the jaw joint.  It is expressed at the articular–quadrate joint in chick, python and fish, and at the malleus–incus joint in mouse.⁵  The same gene marks the same joint.  The joint has simply moved.

Note what kind of evidence that is.  It is not a resemblance judged by eye.  It is a cell-lineage label and a gene-expression pattern in living animals, and neither was available to anyone arguing about this before about 1990.

The part that went the other way, printed for honesty

The developmental data did not confirm everything the paleontologists had inferred.  It killed one of their claims.

Comparative anatomy had proposed that the manubrium of the malleus — the handle that touches the eardrum — was homologous to the retroarticular process of the articular.  Fate mapping showed otherwise.  The processus brevis, not the manubrium, is the second-arch retroarticular homologue, and the manubrium appears to be a novel mammalian structure with no bird or reptile counterpart.⁵

That is worth printing for two reasons.  It shows the developmental evidence is a real test with the power to overturn a paleontological claim, which is why its agreement about the malleus and incus counts for something rather than nothing.  And it is the honest source of the one intelligent-design objection in this area that is not a straw man — see Part 7.

And a straight answer about recapitulation

The objection raised here is that this is Ernst Haeckel warmed over: embryos replaying evolution, an idea discredited a century ago.

No Haeckelian claim is being made.  Nothing above says an embryo passes through its ancestors' adult stages.  The three lines of evidence are a shared condensation, a shared cell lineage, and a shared gene expression pattern — each an observation about a living mouse, chick or python.  Part 7 takes the objection in the form it is usually published, along with the sharper version of it that evolutionists found first.

Part 4 — The anatomy that cannot be assembled by choosing features

Young-earth writers make two different replies to a fossil sequence, and they need different answers.

The first is that the sequence is assembled.  John Woodmorappe's 2001 paper says evolutionary claims about this sequence "typically centre on a relatively small number of features," and that "using isolated bits of evidence, we could construct just about any progression we wanted.  We could, for instance, arrange a sequence of spoons to show a progression in size, thickness, etc."³²  That is a charge about method:  pick the right features and any set of objects will line up.  It is not a claim about burial.  In the same paper Woodmorappe says he has "not disputed" intercontinental correlation or "the temporal succession of mammal-like reptiles."³²

The second is that the order in the rock is burial order.  Kurt Wise's version, given in Part 6, is that these animals lived before the Flood in a belt of habitats between a reptile zone and a mammal zone, and that the Flood buried them in that order.¹  On that view the intermediates are real animals, not descendants of one another.  Nobody on that side claims that sorting built their skulls.

Against the first reply, the double jaw joint is decisive.  Some of these animals have two functioning jaw joints at the same time:  the old quadrate–articular joint and the new dentary–squamosal joint, both present, both articulating, in one skull.  That is not a trend read off a few chosen features.  It is one specific arrangement of bone that is either in the fossil or not, and it sits exactly where the homology from embryos says the change happened.

Against the second reply, the double joint does less, and this article should not pretend otherwise.  It shows the intermediate anatomy is real and that it worked.  A created animal with two jaw joints would show the same.  What the burial-order reply has to explain is why the morphological order and the order in the rock agree, and why the embryos point the same way.  That is the test in Part 1 and the problem Wise himself conceded in Part 6.

Why this particular feature is the one that does the work

Five observations hold together here, and none of them is the kind of thing a choice of features produces.

1.  Seven bones in the lower jaw become one.  The endpoints are not a matter of interpretation.  A basal synapsid jaw has a dentary plus up to seven postdentary bones.  A crown mammal jaw has a dentary.¹⁶

2.  The bones that vanish from the jaw reappear inside the ear, in the same order and the same relative positions.  Articular to malleus, quadrate to incus, angular to ectotympanic.  Not "similar bones" — the same elements, in sequence, with their arrangement preserved.  A trend picked from a few features can be made to run in many directions.  The positions of named bones inside one skull cannot.

3.  There exist animals with both jaw joints at once, working simultaneously.  A. W. Crompton named Diarthrognathus broomi in 1958 — Greek for "double jaw joint" — from the South African Elliot Formation, precisely because it had both.⁶  Morganucodon is the textbook case, described as possessing "a dual jaw joint," with the new joint bearing the load while the old one shifted to hearing, and Morganucodon is the genus with 7,636 catalogued specimens.⁸¹⁰

One correction belongs with point 3.  A 2024 Nature study found that the ictidosaurs, the group that includes Diarthrognathus, reached their dentary–squamosal contact independently of the line that led to mammals, and that some probainognathian clades "independently acquired 'double' craniomandibular contacts."⁷  So Diarthrognathus shows that a double joint works, but it is a parallel case rather than a stage on the way to mammals.  Morganucodon, an early mammaliaform, is the stronger example for that reason.  Part 6 explains why this parallel matters to the young-earth side.

4.  The order of acquisition is constrained and observed.  New joint first.  Load transfers to it.  Old joint reduces and reorients.  Then detachment.  Then the ossified Meckel's cartilage goes.  Liaoconodon is caught at the detachment step with the Meckel's element still ossified and still attached.²²

5.  The same sequence runs every year in living animals.  A newborn marsupial's jaw hinge is its malleus–incus joint until the dentary–squamosal joint takes over.⁹

Lining up spoons by chosen features cannot produce one spoon with two working handles.  That is what these skulls have.

How many double-jointed taxa are there?

This article will not give a number, and the reason is that no published tally exists.

What can be done is naming.  A double jaw joint, or the beginnings of one, has been described in Diarthrognathus, Morganucodon, Kuehneotherium, Sinoconodon, Megazostrodon, Haldanodon, Dianoconodon and Feredocodon, with a partial condition in Riograndia.  That is a list, not a count.  Nobody has published a survey that asks how many synapsid taxa show an unambiguous load-bearing double joint, using a stated criterion, across the whole record.  Anyone who prints a tally is inventing one, and this site is not going to be the source a later writer cites for it.

The list is enough for the argument, because the argument needs the condition to exist in more than one animal and it plainly does.

The strongest form of this evidence is not a fossil at all

A living marsupial runs the transition after birth.

In placental mammals, including humans, the dual joint appears before birth and is gone by the time anyone sees the animal.  In marsupials and monotremes it appears after birth, and until the dentary–squamosal joint forms, the malleus–incus joint is the animal's working jaw hinge.⁹  A newborn opossum suckles on a reptile-grade jaw joint for weeks and then switches over, and the ossicles migrate away into the ear.

The intermediate condition is not a paleontologist's reconstruction.  It is in a pouch in Australia, it has been photographed histologically at every stage, and anyone with a microscope and a licence can check it.

And the intermediate ear worked — this is now quantitative

The other common objection is functional rather than stratigraphic: that an animal midway through this change would chew badly and hear badly, so the intermediates could not have lived.

As of 2025 that is no longer a matter of opinion.  Wilken, Snipes, Ross and Luo built a finite-element model of the mandibular middle ear of Thrinaxodon liorhinus and computed the whole structure's harmonic response to sound.³

Their numbers, in their own terms:

• Hearing range estimated from the allometry of tympanum area and stapedial footplate area: 38 Hz to 1,243 Hz, with a best frequency near 741 Hz or 890 Hz depending on which area is used.
• Hearing range estimated from the finite-element model itself, at 38 dB sound pressure level: 10 Hz to 2,325 Hz when the tympanum alone is stimulated, 2,650 Hz for the whole mandibular ear. Greatest sensitivity around 1,000 Hz.
• The audibility criterion was a stapedial-footplate pressure of 0.0052 Pa, with the hearing threshold set at 28.089 dB by ancestral-state reconstruction.

The result is the part that matters.  Sound applied to the tympanum alone produced enough footplate pressure to hear across the whole estimated range.  Substrate-borne sound conducted through the jaw bones did not meet the threshold at all.

So an Early Triassic cynodont with its ear bones still in its jaw was already an air-borne, eardrum-based listener, comparable in mode to a living mammal, though poor above about 2.5 kHz.  The intermediate configuration is not a hobbled halfway house.  It works as an ear and it works as a jaw, and the model that says so was built by people with no interest in this debate.

Part 5 — How many times did this happen?

This is the genuinely awkward part of the mainstream account.  It is argued out in the open, in Nature, PNAS and National Science Review, and this article treats it as the interesting problem it is rather than as a gotcha.

The problem in one sentence: if the mammalian middle ear is a single hard-to-reach, exquisitely engineered outcome, why does the record keep showing it being reached more than once, in parallel, in separate lineages?

Independence is claimed at four separate levels, and they are not the same claim.

1. The eardrum itself evolved several times

The stem amniote had no tympanic middle ear.  Early stem-amniote fossils have a robust stapes descended from the hyomandibula, doing structural work bracing the skull during biting, with no sign of an airborne-sound role.  Tympanic ears arose independently in the different amniote groups, mostly during the Early Triassic.  Frogs have one too, from a separate origin again.²¹

So "an eardrum with ossicles behind it" is a solution vertebrates found repeatedly and separately.  That is the mainstream position and it is not controversial.

2. The dentary–squamosal jaw contact evolved more than once

This result is new and it is strong.  Rawson and colleagues micro-CT-scanned 25 specimens of three key probainognathian cynodonts and reported two findings that cut in opposite directions.⁷⁷

Riograndia guaibensis, an ictidosaur, independently evolved a dentary–squamosal contact about 17 million years before the character appears in mammaliaforms.⁷

Brasilodon quadrangularis, the sister taxon to Mammaliaformes, does not have an incipient dentary condyle or squamosal glenoid at all, contrary to previous descriptions.  Its jaws articulate solely by the old quadrate–articular joint.⁷

Their conclusion is stated flatly in the abstract: the dentary–squamosal contact "evolved more than once and is more evolutionary labile than previously considered."⁷

3. The ear detached from the jaw at least twice, possibly up to seven times

Ramírez-Chaves and colleagues give the quantitative treatment.¹⁴  They open by naming the problem in the strongest available terms: earlier parsimony analyses of modern and fossil mammals imply "up to seven independent postdentary trough losses or even reversals, which is unexpected given the complexity of these transformations."¹⁴

Their own Bayesian reanalysis, with three-dimensional erosion simulations run to test whether preservation artefacts were driving the result, recovered a simpler answer: the middle-ear bones detach from the postdentary trough exactly twice — once in the ancestors of therians, once in the ancestors of monotremes — with no reversals.  The later step, the loss of Meckel's cartilage, then happened independently in numerous lineages from the Late Jurassic to the Late Cretaceous.¹⁴

Two of their timing figures are useful, cited as the mainstream numbers under discussion and not as this article's evidence: until about 168 Ma every known mammal still had attached middle-ear bones, and every group that diversified from about 163 Ma onward had lost the postdentary trough.¹⁴

Note what kind of number "twice" is.  It is a model output, not an observation.  The seven is what parsimony gave; the two is what a Bayesian model with a different set of assumptions gave.  A future analysis could move it again, and this article should not be read as endorsing either figure.

The most recent count in print is "at least three."  A 2024 Nature paper says the definitive middle ear, the one fully free of the jaw, arose "at least three times independently," citing fossil and developmental studies from 2009 to 2018.⁷  That does not contradict Ramírez-Chaves's "twice," because the two figures count different steps.  Their two is the loss of the trough on the dentary.  The final release from the ossified Meckel's cartilage came later and, on their own reading, many times over.  So the published range runs from two to seven, depending on which step is counted and which model counts it, and three is what the newest paper treats as the floor for full separation.

4. Monotremes against therians — a live dispute about millimetre-scale bones

The original claim is Rich and colleagues in Science in 2005, arguing for independent origins of the middle-ear bones in monotremes and therians.  It rested on Teinolophos trusleri, an Early Cretaceous stem monotreme from Victoria, and specifically on a groove in its dentary read as a Meckelian groove — implying the ear was still attached in a monotreme relative long after therians had detached theirs.³⁰

It has not settled.  Two 2021 papers in National Science Review argue it out.  Meng and Mao contend that a partly overlapping "braced hinge" incudomallear joint is primitive for Mammalia, requiring four evolutionary steps, against a monotreme-like fully overlapping joint that they count as requiring at least five.³³  Wible and Shelley reply that Meng and Mao lumped five separately-scored characters into two and then optimised the lump, and that the tree attributed to their opponents was never an analysis those authors ran.³⁴

What the dispute is actually about is instructive.  Meng and Mao complain that a CT voxel size of 32.7 µm may not be fine enough to tell a malleus from an incus in the specimen at issue, and that the digital data were not released for them to check.  This is a real methodological argument between working paleontologists, conducted in public, about sub-millimetre bones in single fossils.  It is not a field closing ranks.

What this does to each side's argument

It does not count against the transition, and a creationist should not pretend it does.  Every paper cited above assumes common descent and is arguing about topology — about how many times, not whether.  Rawson's Riograndia result in particular makes the dentary–squamosal joint look easier to evolve, not harder, because it shows a cynodont reaching it 17 million years early without being anybody's ancestor.

Where it does bite is on irreducible complexity.  A structure claimed to be unreachable by gradual steps is a poor candidate for that claim if the record shows several lineages reaching parts of it independently.  If the mammalian middle ear is an all-or-nothing design, it is an odd all-or-nothing design that keeps getting partly assembled by animals that are not mammals.

The mainstream has a candidate mechanism for the repeats, and it has been tested in living animals.  Anthwal and colleagues blocked the cells that dissolve cartilage, in mice by knocking out the gene c-Fos or with the drug alendronate, and in opossums with alendronate.  Meckel's cartilage then stayed in place, and in the mice it ossified and left a groove on the dentary.  In their words, "This morphology phenocopies the ossified MC and Meckelian groove observed in Mesozoic mammals."  They propose that a shift in the timing of those cells was a key step to the free ear.³⁵³⁵  If one change of timing separates the transitional ear from the free one, the same release happening in several lineages is not surprising.  The proposal is a hypothesis about extinct animals, and the authors say the signal that changed is unknown.³⁵

Where it helps the young-earth side is narrower, and it is real.  Several separate groups becoming more mammal-like in the same stretch of rock is exactly the observation Kurt Wise built his 2006 argument on, and the 2024 paper confirms it for the jaw joint in one of the families he named.  Part 6 gives his argument and its limits.

Where it bites on the mainstream is the parsimony cost, and they name it themselves.  Seven independent losses of the same complex arrangement is a lot, and Ramírez-Chaves and colleagues call it unexpected in their own abstract.  Their reanalysis reduces it, but by changing the model, which is a legitimate move and also a move a critic is entitled to notice.

And for this series' running theme, it is a clean case.  Morphology alone gave an unstable answer here.  The correction did not come from re-eyeballing the same bones — it came from better scans, more specimens and explicit models.  That is the same lesson the convergence section of the flagship article draws, and it is a lesson about the limits of morphological inference, not about whether animals are related.³⁶

Part 6 — What a young-earth creationist paleontologist said about this

Kurt Wise holds a PhD in paleontology from Harvard.  He is a young-earth creationist.  In 1995 he published an assessment of what the fossil record does and does not support, in a creationist journal.¹

The credential is printed in the paper itself, in the author box on its last page: a B.A. from the University of Chicago, and an M.A. and Ph.D. in palaeontology from Harvard.¹  The paper does not name Gould.  That Wise studied under Gould at Harvard is printed in the author note to his own chapter in the creationist anthology In Six Days (2001), and the historian Ronald Numbers calls him the "protégé of the archcritic of creation science, Stephen Jay Gould," who "headed east to work with Gould at Harvard."^37373838

His vocabulary, which the quotation cannot be read without

Wise builds a four-term vocabulary before he concedes anything, and the concession is stated in those terms.¹

A stratigraphic intermediate lies between two fossils or two fossil groups in the rock.  A morphological intermediate is anatomically between them.  A stratomorphic intermediate is both at once — the only one of the three that is evidence for anything.

He then says macroevolution predicts four kinds of stratomorphic intermediate, and rules on all four.  The first fails.  Intermediates between species have been, in his words, "the most disappointing for classical Darwinists," and their absence is what drove the acceptance of punctuated equilibrium.¹  The other three, he reports, are met — intermediate species, intermediate higher taxa, and stratomorphic series.

Anyone quoting the concession without that sentence is misquoting him.  His accounting is one out of four failed, three out of four met.  Quoting only the concession overstates it.  Quoting only the "most disappointing" line understates it.  Both belong, in that order.

The concession

Under his own section heading "It Is a Very Good Evolutionary Argument," having listed the mammal-like reptiles among the higher-taxon intermediates, he writes:

"Creationists therefore need to accept this fact.  It certainly CANNOT be said that traditional creation theory expected (predicted) any of these fossil finds."¹

"CANNOT" is in full capitals in the original.

The second concession, which is the one that belongs in this article

Most citations of Wise stop there.  He says something far more specific about this sequence two pages later, under a heading arguing that the record actually encourages the creationist model.

He writes that the mammal-like reptiles lie stratigraphically between the oldest reptiles and the oldest mammals, making them a good higher-taxon stratomorphic intermediate.  He then adds that as one rises stratigraphically through them, their teeth show "a virtually monotonic change" from cone-shaped reptilian teeth toward complex mammalian ones — and that the three jaw bones show a virtually monotonic change from three subequal bones toward the single bone of the mammal jaw.  He concludes that they are therefore morphological intermediates as well as stratigraphic ones.¹

His own authority for the jaw claim is not a creationist.  His footnote points to J. A. Hopson's 1994 chapter on synapsid evolution in the Paleontological Society's Major Features of Vertebrate Evolution.¹

That paragraph is a young-earth creationist with a Harvard paleontology PhD stating, in a creationist journal, exactly the claim this article is about.

One qualification has to run with it, and it comes from the mainstream side.  Wise's word is "monotonic," and the 2023 phylogenetic test in Part 1 found that the postdentary-reduction measure is not monotonic across non-mammalian synapsids as a whole.¹⁵  So Wise conceded slightly more than the data support, in the direction against his own position.  The trend is real in dentary area and real in the cynodont-to-mammal segment; it is not a smooth ramp running the whole length of the record.  That correction is offered here because the alternative is quoting a creationist's overstatement as though it settled something.

What Wise does next, which this article may not omit

Wise does not stop at the concession.  The rest of that page proposes a creationist alternative, and quoting the concession without it would be exactly the kind of selective use this site audits other people for.

His argument, in paraphrase: the cynodont groups are distinguished mainly by their teeth; different teeth imply different diets; diets intermediate between reptile and mammal diets imply the animals lived in a geographic belt between a reptile zone and a mammal zone before the Flood; and a transgressing flood would then bury them in the intermediate stratigraphic position where they are found.  He predicts that several independent mammal-like-reptile lineages, herbivores and carnivores separately, should each become more mammal-like up-section, and asserts that "this is exactly what is seen."  He argues that conventional theory has to explain the same pattern by parallel and convergent evolution, and suggests the creation model may turn out to be the better explanation.¹

His conclusion concedes and then hopes.  Substantial supporting evidence for macroevolutionary theory can be found in the record of stratomorphic intermediates, he writes, and the creation model "is not well enough developed at present" to evaluate that evidence or to offer an adequate alternative.¹

That was 1995.

Wise's 2006 restatement, and the part of it that has held up

Wise came back to the argument once, in 2006, and sharpened it.  In a plenary abstract to the Baraminology Study Group he proposed a test for sorting fossil sequences into two kinds:  change within a created kind after the Flood, and burial of pre-Flood zones during it.  The horse series passed his test for change within a kind, because it is one family, it follows the rocks, and it tracks a change in vegetation and climate.  The synapsids fell in the other class.  He wrote that "a number of reptile families show an increase in mammal-likeness in Triassic sediments (cynognathids, tritylodontids, trithelodontids, chiniquidontids, traversodontids, diademodontids)," and that because these sequences "lack an explanatory environmental trend," the convergences "may be due to transitional ecologies in the pre-Flood world rather than intrabaraminic diversification in the post-Flood world."¹⁷¹⁷  He was candid about the method:  explanations of fossil sequences "have thus far been done in a post hoc fashion and run the risk of becoming ad hoc in nature."¹⁷

The observation in that argument has held up.  Wise's point is that mammal-likeness rises in parallel in several separate families, not along one line.  In 2024 Rawson and colleagues found that one of the families he named, the tritheledontids, together with their relative Riograndia, reached the dentary–squamosal jaw contact on their own, and that several probainognathian groups "independently acquired 'double' craniomandibular contacts."⁷  The same paper counts at least three separate origins of the free middle ear.⁷  A second short-age scientist makes the same observation.  Leonard Brand, a biologist, writing with the geologist Arthur Chadwick, grants that the Mesozoic rocks "hold series of organisms that can be interpreted as a good evolutionary sequence from reptiles to mammals," and adds that "In the Triassic deposits, several groups simultaneously show more mammal-like traits."³⁹³⁹  The parallels are real and the mainstream now prints them.  Wise pointed to them in 1995 and 2006, before the jaw-joint result was published, and the result fits what he described.

What the observation does not show.  Parallel change is expected under common descent too.  Close relatives share the same developmental machinery, and Part 5 gives a tested mechanism by which one change of timing frees the ear from the jaw.³⁵  The parallel families are not scattered across the tree:  all six that Wise names are eucynodonts, nested inside the same tree whose order matches the rock in Part 1.  And a zonation model ties mammal-likeness to diet and habitat, which can speak to teeth and jaw muscles but gives no reason why a feeding zone should move bones into the ear.

Brand states the difficulty more plainly than Wise.  He writes that "the confusing group of Triassic fossils with apparently intermediate structures is a puzzle," and offers only a suggestion:  "perhaps the same genes form lower jaw bones or middle ear bones, dependent on epigenetic factors, and in these Triassic forms, there were epigenetic aberrations."³⁹  The nearest experiment is the mouse work in Part 5, where blocking one cell type left a mouse with a persistent, bony Meckel's cartilage and a jaw groove like a Mesozoic mammal's.  That shows the developmental switch exists, which Brand's suggestion needs.  It does not produce a double jaw joint, it came from a mutation or a drug rather than an epigenetic change, and its authors read it as "a partial recapitulation of an ancestral form."³⁵  It does not touch the order in the rock, which is the part Wise conceded.

So the alternative stands where Wise left it in 2006:  an observation that holds, attached to an explanation with no map.  Thirty-one years after the first sketch there are still no zone boundaries, no measurements, no quantitative predictions and no stated conditions under which it would fail.

The geography has to be drawn, though it is not impossible.  The synapsid record is not a single band in one place.  It runs through the Karoo Basin, the Russian platform, China, Argentina, Brazil and Texas, with the same trend appearing in each in the same order.²⁰  On the mainstream reconstruction those regions were parts of one supercontinent, Pangaea, when the beds were laid down, and therapsids moved between them:  Kammerer and Masyutin write of "the extensive intercontinental dispersal inferred for coeval therapsid groups," though they also found gorgonopsians split into a Russian clade and an African one.²³  So a single belt of zones around one landmass is not ruled out.  What the model owes is the map:  where the belt lay, and why each basin received the zones in the same order.  Neither the 1995 paper nor the 2006 abstract supplies it.

One note on the quotation, because this site audits its own sources

Our own research notes had it slightly wrong.  They dropped the word "be" from "It certainly CANNOT be said," and misspelled phenacodontids.⁴⁰  Small errors, but the whole point of checking is that small errors compound, and this one changes the sentence into something Wise did not write.  Every quotation in this article was transcribed from 300-dpi renders of the printed journal pages, read as images, because that PDF has no text layer.¹

Three cautions about using Wise as a witness

If this article is going to lean on him, a reader is entitled to the whole picture.

On the facing page of the same paper, Wise suggests that Loch Ness sightings may be living Basilosaurus.¹  That is in the 1995 article, a few inches from the concession being quoted.  It does not make the concession false — a concession against interest is if anything strengthened by coming from someone with unconventional views — but anyone citing this paper should know what else is in it, and should not be surprised by it later.  Creation Ministries International has since placed living plesiosaurs among the arguments creationists should not use.⁴¹

His 2002 book restates the creationist case far more confidently and does not mention the 1995 concession.⁴²  Whether that is a change of mind, a change of audience, or an omission is not something this article can settle.  It is recorded because quoting the 1995 paper without noting the 2002 book would be selective.

His most important later work concedes more, not less.  In a 2023 conference paper with Donna Richardson, Wise ran an original analysis of 114 stratigraphic boundaries using the same public database this article uses, counting species that occur on both sides of each boundary within two degrees of latitude and longitude.  Two of his findings cut against his own movement, and he printed them: "the global biostratigraphic column seems to be real," and that burial order bore no relationship to fluctuations in flood energy.⁴³  That is a young-earth paleontologist doing original quantitative work and reporting a result against his own model — which is the standard this site is trying to meet.

That is the honest state of play.  The best-credentialled young-earth paleontologist of his generation looked at this evidence, conceded it, offered an alternative, never built the alternative out, and has since conceded more.

One more thing he wrote belongs here.  On the same page as his vocabulary, Wise tells creationists — in capitals in the original — not to use the phrase "transitional form" in their argumentation, because the term is interpretive and ambiguous.  On the same page he says that Gould's different definition of the term has led some creationists to conclude wrongly that Gould does not regard Archaeopteryx as transitional.¹  That is a trained creationist paleontologist telling his own movement, in 1995, that it misreads its favourite hostile witness.

Part 7 — The objections, one at a time

Five creationist sources on this sequence were read in full for this article, and one negative finding is worth stating first.

Intelligent design has essentially nothing on it.  Searches of the Discovery Institute's news outlet return no substantial article on the mammalian middle ear.  The sequence is not among the ten icons Jonathan Wells attacks in Icons of Evolution — those are the Miller–Urey experiment, Darwin's tree of life, vertebrate limb homology, Haeckel's embryos, Archaeopteryx, peppered moths, Darwin's finches, four-winged fruit flies, fossil horses and human ancestry — and it is not among the ten he revisits in Zombie Science.⁴⁴  The best-documented transitional sequence in the vertebrate record is the one the movement's most prominent critic of "icons" did not take on.

"There are no transitional forms between reptiles and mammals."

Dead.

Refuted in a creationist journal in 1995 by Wise.¹  Refuted by 809 named non-mammalian synapsid genera.¹²  Refuted by the double-jointed taxa.  Creation Ministries International's own list of arguments creationists should not use kills the flat form of this claim outright.⁴¹

Anyone still making it has not read their own side's best paleontologist.

"The majority of characters reverse rather than progress up the sequence."

This is John Woodmorappe's 2001 argument, and it is the most quantitative objection in the creationist literature on this subject.  It deserves to be described properly before it is answered.³²

What he actually did.  He took published cladistic character matrices, scored each character as "progressive" — changing once, toward the mammalian state — or "reversing" — changing direction at least once, summed the scores per taxon, and asked whether the totals climb smoothly.  His three datasets and his own counts:

His headline sentence is that "the majority of anatomical traits… do not show a unidirectional progression towards the mammalian condition!"  His conclusion is that the mammal-like reptiles cannot, "by any stretch of the imagination, be portrayed as some sort of quasi-lineage."³²

Verdict: the objection fails, for five reasons, and the first is the one our own side should be pointing out.

1.  His headline claim is contradicted by his own largest table.  In the Sidor and Hopson dataset — his primary dataset, and by far his biggest at 165 usable characters — 88 are progressive against 77 reversing.  Progressive characters are the majority, at 53%.  His claim holds only in his two smaller datasets.  That is an arithmetic error a reader can check inside his own article, in a table he printed himself, and it is the strongest single objection in creationist writing on this subject failing on its author's own numbers.

2.  He treats a cladogram as a ladder while admitting it is not one.  He grants in his own text that cladograms are not intended to identify ancestor–descendant relationships, and then requires the taxa to form a smooth linear progression anyway.  Side branches — tritylodontids, tritheledontids, dicynodonts — will always show "reversals" relative to an arbitrary linear ordering, because they are not on the line.  A large share of the reversals he counts are an artefact of the ordering he imposed.

3.  He never computes a null.  Homoplasy is expected under evolution and is routinely quantified: every published matrix has a consistency index and a retention index measuring exactly this.  "77 of 165 characters reverse somewhere" is not a result until somebody says what a real evolutionary tree produces.  Woodmorappe reports no consistency index, no retention index, and no simulation.  A number with no expectation attached cannot falsify anything.

4.  Summing binary characters produces a dimensionless score, so "gap size" in it has no meaning.  Adding up presence/absence characters gives a count, not a morphological distance.  Whether a jump from 54 to 104 is "big" depends entirely on how many characters happen to have been coded in that region of the tree, which is a function of collecting effort and of how much of each animal survives.

5.  The matrix he used says the opposite, and he does not engage with what its authors concluded.  Sidor and Hopson's paper exists specifically to test gradualism in this sequence, and its answer is that "in no case was a long ghost lineage associated with few apomorphies," concluding that "gradual character evolution predominated."²⁸  Woodmorappe reused their characters and reached the opposite verdict without addressing their test.

What survives from Woodmorappe, and it is not nothing.  His observation that tritylodontids are in some respects more derived than the earliest mammals is a real feature of the data, and he sourced it to real quotations from mainstream workers.  He was right that the sister-group question was unresolved in 2001.  And his general point — that mammalian characters appear in a mosaic rather than in lockstep — is now the mainstream position, confirmed in Nature twenty-three years later by the paper that found the dentary–squamosal joint evolving more than once.⁷  The 2023 phylogenetic test of dentary proportions points the same way.¹⁵  Cite him for the mosaic.  Do not cite him for the demolition.

What is dead in Woodmorappe.  His single paragraph on the jaw-and-ear transformation itself — the actual subject — rests on one quotation saying the intermediate stages in the transfer of the postdentary elements are poorly documented, with the only evidence being muscle attachment sites in Amphitherium and Peramus.  That was arguably fair when he wrote it.  It is now obsolete.  Yanoconodon in 2007, Liaoconodon in 2011, and Origolestes, Dianoconodon and Feredocodon in 2024 are precisely those intermediate stages, with the bones preserved in three dimensions.²²⁸

"The intermediate animals could not have heard. The bones would have had to cross the eardrum."

Two errors, and the second is the interesting one.⁴⁵

The first is answered by measurement.  Wilken and colleagues modelled the intermediate ear in 2025 and it hears — better through the eardrum than through the jaw, which is the opposite of what the objection assumes.³

The second is that there was no eardrum to cross.  The stem amniote had no tympanic ear at all.²¹  The mammalian eardrum developed in a cleft on the outer face of the angular bone, alongside the postdentary bones rather than across them.  The picture the objection has in mind — bones migrating past a pre-existing reptile eardrum — describes a situation that never existed.

"Why is there no animal with two ear bones, halfway between one and three?"

The question is malformed.⁴⁵

The stapes was already an ear bone before any of this began.  The count runs one to three, and the intermediate stage is a three-ossicle ear still attached to the jaw.  There was never a two-ossicle stage to look for.

Morganucodon is the animal the question asks for, and there are 7,636 catalogued records of it.¹⁰

"Fate mapping shows the malleus is not purely first-arch, so the homology is broken."

Real, much smaller than claimed, and the mainstream literature reported it first.⁴⁵

The finding is genuine: Hoxb1 fate mapping shows the processus brevis of the malleus is second-arch derived, and it overturned the older paleontological claim that the manubrium was the retroarticular homologue.  The manubrium turns out to be a novel mammalian structure.⁵

But that correction does not touch the malleus/articular or incus/quadrate homology, which rests on the shared condensation, the first-arch origin of the main body of each bone, and the shared Bapx1 joint expression.⁵⁵⁵

A creationist citing the dual-arch problem should also cite that the people who found it are the same people who reaffirmed the main homology, in the same papers.  Using a paper's correction while ignoring its conclusion is the practice this site criticises when it is done to us.

"The ossified Meckel's cartilage in Liaoconodon was assumed, not preserved."

Factually wrong, and worth correcting because it appeared on a major creationist site.

The claim, verbatim, is that "Though not preserved in the fossil record, the cartilage is assumed by these researchers to have been there, holding the migrating bones in place."⁴⁶

One sentence of the paper being discussed settles it.  The ectotympanic and malleus "have lost their direct contact with the dentary bone but still connect the ossified Meckel's cartilage."²²  The element is ossified.  It is bone.  It is preserved, and it is the reason the paper was published.  The mistake appears to be a confusion with the unossified Meckel's cartilage of living mammals.

The same page gets something else right, and both facts should be printed together.  It quotes the paper accurately as saying that "fossils illuminating this critical transference are sparse and often ambiguous."⁴⁶  That sentence is in the published abstract.  So the article quoted the paper correctly on its own admission of sparseness and then got the central fact about the specimen wrong.  Correcting one does not entitle anyone to ignore the other.

"These conclusions depend on the discredited idea that embryos replay evolution."

A straw man of a real issue.⁴⁶

No Haeckelian claim is made anywhere in this literature.  The developmental evidence is about shared cell lineages, shared cartilaginous condensations and shared gene expression, not about ontogeny repeating phylogeny.  And the observation about newborn marsupials is an observation, not an assumption: the ossicles serve as the jaw hinge after birth and then detach, and it has been photographed.⁹

The honest version of this problem has already been published by evolutionists, and it is more interesting.  Loss of Meckel's cartilage during development does appear to mirror the fossil sequence.  Loss of the postdentary trough does not.¹⁴  That is a partial mismatch between development and the record, found and printed by the side arguing for the transition.  A creationist who wants a real argument here should use theirs.

"Liaoconodon is far too late to be a transitional form."

This is the best of the creationist objections and it deserves a direct answer rather than a wave.⁴⁷

The stratigraphic fact is correct and this article concedes it.  Liaoconodon is Early Cretaceous, around 120 Ma on the mainstream scale.  Definitive mammalian middle ears are tens of millions of years older.  Nobody disputes that.

The inference does not follow, because nobody claims Liaoconodon is an ancestor.  It is offered as a lineage that retained an intermediate condition long after the transition had happened elsewhere.  A living coelacanth illustrates lobe-fin anatomy without being anyone's ancestor, and a living platypus shows early mammal features without being anyone's ancestor either.  Doyle's argument works only against a claim the paper does not make.

His homoplasy point, though, is conceded by the field, loudly.  He objects that the evolutionary reading requires numerous independent losses.  It does, and Ramírez-Chaves and colleagues open their own paper by saying so and calling it unexpected.¹⁴  Doyle is pointing at a real cost.  Where his argument becomes unfair is in treating an acknowledged cost as a concealed one.

"Fossils from the Jehol deposits deserve automatic suspicion because of Archaeoraptor."

Our own side should stop making this move.

Doyle raises the Archaeoraptor forgery in connection with the Jehol biota, writes that he is not suggesting any particular find from there is fraudulent, and then says there is reason for a priori scepticism about fossils from the region.⁴⁷

That is insinuation rather than argument.  Archaeoraptor was exposed by paleontologists, not by creationists.  It was never published in a peer-reviewed journal.  And a young-earth article that wants to be believed cannot ask for charity about its own sources while denying it to everyone else's.

"It is just a series of similar animals arranged by an evolutionist."

Two objections share this wording, and Part 4 separates them.  If it means the series was built by choosing a few features, as in Woodmorappe's spoon analogy, the double jaw joint answers it:  no choice of features gives a skull with two working jaw hinges.³²  If it means the order is burial order, the double joint does not answer it, because a created animal could have two jaw joints too.  That form is met by the order test in Part 1 and by Wise's own concession in Part 6.  As an objection it does not work today, because nobody has built the model it needs.

Part 8 — Errors on both sides

A test of whether an audit is honest is whether it corrects the side it favours.  This one found four things wrong in the mainstream literature, three of them still in circulation.

"Probainognathus is a Late Jurassic cynodont."

It is Middle-to-Late Triassic — Ladinian to Carnian, from the Chañares Formation of Argentina, named by Romer in 1970.⁵

The published claim is off by roughly ninety million years, and it is in a heavily cited review in the Journal of Anatomy.  It does not affect that paper's conclusions.  It is exactly the sort of thing that gets copied forward by people who do not check, and this article is where it stops.

The same paper mangles Reichert's date twice — printing "Reichert, 1937" in the text and a reference-list entry reading "1937;1837:120–220" — and dates Gaupp to 1912 rather than 1913.⁵

"Brasilodon has an incipient dentary condyle and squamosal glenoid."

Reported for about twenty years and repeated widely, because it made Brasilodon the taxon closest to mammals and gave the transformation a tidy last step.

Micro-CT scanning in 2024 showed it has neither, and that its jaws articulate solely by the old quadrate–articular joint.⁷

This is the single largest correction in the recent literature on this sequence, and it went against the story rather than for it.  A reader is entitled to ask how many of the remaining reconstructions will move the same way, and this article does not know the answer.

"Tikitherium copei is the oldest mammaliaform, at 237 million years."

The specimen "actually represents the last upper premolar (P4) of a Neogene crocidurine soricid insectivore," which is to say a shrew tooth, tens of millions of years too young.²⁶

The Paleobiology Database still carries the old assignment, and its neighbour from the same collection, Gondwanadon tapani, which is why that database gives Mammaliaformes a 237 Ma first appearance and why the table in Part 1 leaves the clade out.  A database is a compilation of the published literature and inherits its mistakes; a first-appearance date pulled from one without checking the underlying record is a filing decision, not a fact.

And note which way this cuts: an apparently out-of-place fossil that a creationist could have used was removed by mainstream paleontologists, not by us.

And the correction that cuts against the mainstream's own storytelling

The adaptive explanation everyone repeats is wrong.

For decades the textbook account was that consolidating the jaw into one bone made the skull stronger and the bite more efficient.  In 2023 Lautenschlager and colleagues tested it with digital reconstruction and biomechanical modelling across Thrinaxodon, Diademodon, Exaeretodon, Probainognathus, Morganucodon, Hadrocodium and the living opossum Monodelphis.  They found "no evidence for an increase in cranial strength and biomechanical performance."¹⁶  Stresses fell in the braincase and skull roof but rose in the zygomatic region, and mechanical advantage decreased.  Their replacement explanation is miniaturisation plus a dietary shift to insectivory.

The transformation happened.  The tidy reason given for why it happened does not survive testing.

Two consequences.  Anyone still saying "one jawbone is stronger" is repeating something that has been tested and failed.  And Wise's 1995 aside is the same claim — he suggested that complex mammalian teeth may require the increased jaw strength that one bone gives over three — so it did not survive either.¹

What would change this verdict

Stated as conditions, so that a reader can hold this article to them.

• A young-earth model of pre-Flood ecological zonation developed to the point of making measurable predictions about which synapsids should occur where, with stated failure conditions. Wise proposed this in 1995 and nobody has built it.¹
• A demonstration that the double-jointed taxa are misinterpreted — that the second joint is not load-bearing, or is an artifact of preservation or of reconstruction. That would have to survive the Dianoconodon and Feredocodon geometry, which is where the load argument is strongest.⁸⁸
• A stratigraphic inversion in the clade run in Part 1 larger than the dating brackets, established from superposition rather than from a date. The one reversal found so far is smaller than the brackets.²⁰
• A count of the coordinated genetic changes required, which would make a waiting-time argument possible for the first time. Until that number exists, this article holds that the argument is not available here.
• A published tally of unambiguously double-jointed taxa, using a stated criterion. There is not one. It would strengthen the mainstream case, not weaken it, and it is odd that nobody has done it.

Sources

The research dossier behind this article, with the full specimen census, the query log, the Wise audit and every objection assessed, is published alongside it:  research/synapsid-sequence.md.

The quotation audit for this series — 236 passages checked against sources actually opened, with every verdict and every failure, including the ones that went against our own side — is a companion page.⁴⁰