The Horse Series

"Fossil Equidae:  a monobaraminic, stratomorphic series." — the title of the most-cited creationist technical paper on fossil horses.  Its three young-earth authors argue that the fossil horses are one created kind, and that they appear in the rocks in a real order.¹

The ladder diagram "is much more than merely wrong.  It never could provide the promised illustration of evolution progressive and triumphant — for it could only be applied to unsuccessful lineages.  Bushes represent the proper topology of evolution." — Stephen Jay Gould, 1991.  Creationists quote the first half.  The clause after the dash, and the sentence after that, are almost always cut.²

Summary

The row of horses in the museum case is the most famous fossil sequence in the world.  Small four-toed animal on the left, Equus on the right, five or six steps in between.  It has been printed in textbooks since 1903.

It is wrong, and paleontologists demolished it decades before creationists took it up.

That is the first thing this article says.  Here is the second, and it is the one that decides what the article is about.

The horse series does not test the boundary creationism defends.  Every animal in it — the nine on the museum ladder, the fifty-one genera in the real record, all of it — sits inside one family, Equidae.  Young-earth biology already grants common ancestry at that rank.  That is what a created kind is.  The most-cited creationist technical paper on the subject says so in its title.¹

But the concession is not free, and it should not be described as free.  That same paper found, by its own statistical test, that the earliest horses and the modern ones differ the way two separate kinds would, and that only the fossils in between join them.¹  Its authors place the whole series after the Flood, which means the change from small four-toed browsers to large one-toed grazers "must have began and subsided rapidly."¹  The Institute for Creation Research's current model places most of the same rocks inside the Flood instead.³  So the horse record does test young-earth models, on two points the family rank does not settle:  how fast a created kind can change, and where the Flood ends in the rock.  The section "What the concession costs" gives both.

A creationist who attacks the horse record is therefore defending ground his own movement's paleontologists conceded thirty years ago, with credibility he needs somewhere else.  The honest fight is over the rate and the timing, not over the fossils.

So what is this article about?

Honesty of presentation.  Not whether horses evolved, which both sides answer the same way at this rank, but how a well-sampled fossil sequence has actually been described to the public.

On that question both sides have a case to answer, and this article prints both.

What this article concludes:

• The ladder icon is wrong. Equid phylogeny is a bush of about 51 accepted genera in which "ancestors" outlive "descendants," species stay unchanged for millions of years, and roughly a fifth of the size lineages run backwards.
• Mainstream paleontology said so first, and said it hardest. George Gaylord Simpson attacked the linear reading in 1951. Donald Prothero and Neil Shubin found stasis and abrupt branching in the classic Mesohippus–Miohippus transition in 1989. Gould devoted an essay and a chapter to calling the icon a joke. None of them was a creationist and none of them concluded that horses did not evolve.
• The animal at the foot of the icon is not currently classified as a horse. Froehlich's 2002 revision restricted Hyracotherium to its English type species and called it a primitive equoid, outside Equidae. The Paleobiology Database files it in Palaeotheriidae today.⁴ Twenty-four years later, the museums have not caught up.
• The creationist account of how that happened is backwards on every point. Eohippus was dropped in the mid-twentieth century on nomenclatural priority — Owen's 1841 name beats Marsh's 1876 name — not because anyone decided the animal was not a horse. And when the demotion did come, in 2002, it resurrected Eohippus rather than retiring it.
• The record is dense, not thin. 6,517 occurrence records, 51 accepted genera, and 288 accepted species out of 440 names ever proposed. A third of the names coined for fossil horses have been sunk as synonyms. You cannot over-split a record you do not have.
• The stratigraphic order is real, and it does not depend on any date. Small, low-crowned, many-toed forms occur low. Large, high-crowned, one-toed forms occur high. That rests on superposition within each basin and on matching fossils between basins, and young-earth geologists accept and use both. The creationist baraminology paper ranks its own horses against that order.¹
• The trends have real reversals, and the mainstream found them. The earliest horse of all, Sifrhippus, got smaller through a sharp warming event at the very bottom of the ladder, an interval the mainstream dates at 130,000 years.⁵
• Young-earth writers disagree with each other about what the horse fossils are. Cavanaugh, Wood and Wise read them as diversification after the Flood. Tomkins and Clarey of the Institute for Creation Research put Tertiary mammals in the last Flood layers. Jonathan Sarfati doubts the trend and doubts that the first animal was the ancestral horse. "What the concession costs" sets them side by side, with the weak point of each.
• The newest creationist treatment of this topic fails an audit. Jerry Bergman's 2024 paper in Creation Research Society Quarterly contains six sound claims and a long list of bad ones, including two flat falsehoods, an inverted reading of a live anatomical dispute, a misattribution, a misdating, and a quotation routed through a 1971 creationist secondary whose reference list dates the same book two different ways. Part 7 gives all twenty with verdicts.
• One of his claims was checked against the database and the truth turned out to be better than the claim. Part 5.

The rank question, stated up front

This has to come before anything else, because it decides how much the rest of the article is worth.

Young-earth biology accepts common ancestry within a created kind — usually somewhere around the family level.  A fossil series only counts against that view if it crosses a higher boundary.  The mammal-like reptiles cross a class boundary.  The whales cross an order boundary.  Fish to tetrapods crosses a class boundary.

The horse series crosses nothing.  It is Equidae from end to end.

And creationists have said so themselves, in print, repeatedly.

• Kurt Wise, 1995. Wise took his doctorate in paleontology at Harvard, where he studied under Gould.⁶⁶⁷⁷ Writing in a creationist journal, he listed the horse, camel, elephant, pig and titanothere sequences among the fossil "stratomorphic series" that he said creationists have to accept.⁸ A companion audit for this series sorted his nine examples by the rank of the gap each one bridges. Horses, camels, elephants, pigs and titanotheres are all within-family or within-superfamily, so none of them crosses the boundary young-earth creationism defends. The expensive entries on Wise's list are elsewhere.⁹
• Wise, 2002. In Faith, Form, and Time he wrote that the Cenozoic sequences — horses, elephants, camels — may well be stratomorphic series documenting diversification within a created kind.¹⁰
• Cavanaugh, Wood and Wise, 2003. Their paper for the Fifth International Conference on Creationism is titled "Fossil Equidae: a monobaraminic, stratomorphic series." They scored nineteen fossil horse species, from Hyracotherium to Equus, on 33 anatomical characters taken from a 1989 mainstream dataset, and concluded that all nineteen belong to one created kind. They then ranked the species by shape and found that the ranking follows the order in which the species first appear in the rock, with a correlation above 0.87. Their reading of it: "a rapid, post-Flood, intrabaraminic diversification."¹¹

One caution before leaning on that concession.  The method behind the 2003 paper is statistical baraminology, and in 2021 a creationist statistician argued in Answers in Genesis's own journal that the technique should be abandoned.⁹  The authors were careful too.  Their characters are 64% teeth, they sampled nineteen of about 150 species, and they call their findings "preliminary."¹  Report the paper as a published creationist position, not as a measurement.  It is cited here for what its authors concluded, not for the strength of the tool they used.

Consequence, stated plainly.  Nothing in this article bears on the origin of Perissodactyla, of mammals, or of anything else creationism disputes.  The gap creationism actually cares about lies below Equidae — where the odd-toed hoofed mammals came from at all — and that question is not researched here, in either direction.¹¹

What the concession costs

Granting that the fossil horses are one kind does not end the argument.  It moves it.  Three costs come with the concession, and young-earth writers have not yet paid all of them.

First, the two ends of the series do not look like one kind.  The 2003 paper's test compares every pair of species.  Every early hyracothere came out significantly unlike every member of the modern horse subfamily, which the method normally reads as a boundary between kinds.  The authors write:  "Normally, we would interpret this as evidence of two different equid holobaramins, but four fossil equid species prevent this conclusion."¹  The four are Mesohippus, Miohippus, Archaeohippus and Parahippus, which are the intermediates.  The authors argue that a long, thin cloud of species will always show its two ends as unlike, so the result still fits one kind, and that is a fair point about the statistics.  A skeptic will answer that a chain of intermediates joining two very different end forms is what a transitional series looks like, and that the method gives no rule for how long such a chain may grow before it stops being one kind.  That answer has force, and no young-earth paper read for this article has met it.

Second, the change has to be fast.  Cavanaugh, Wood and Wise place the whole series after the Flood and say the diversification "must have began and subsided rapidly."¹  Whitmore and Wise place the Green River beds of Wyoming, which hold Hyracotherium, at "decades to several centuries following the Flood," and suggest the small early forms bred fast:  "Such species probably had shorter generation times and larger litters."¹²¹²  On that reading, everything from a small four-toed browser to a large one-toed grazer with tall, cement-covered grinding teeth fits into the centuries after the Flood.  Size alone is easy, since breeders have made both the Shire horse and the Falabella.  Losing the side toes and building a grazer's teeth is a much larger change.  Mechanisms have been proposed.  Jonathan Sarfati appeals to created genetic information that was "latent" and later sorted by selection.¹³¹³  Todd Wood proposed mobile "Altruistic Genetic Elements" as "a solution to the problem of rapid post-Flood diversification of baramins."¹⁴¹⁴  Neither has been tested against the horse record.  This is the weakest point of the young-earth reading of horses, and it should be presented as an open problem, not as a solved one.

Third, it needs the Flood to end low in the rock, and young-earth geologists disagree about that.  The 2003 reading holds, in the authors' words, "assuming that the Flood/post-Flood boundary occurs approximately at the Cretaceous/Tertiary (KT) boundary," and they note "considerable dispute among creationists" over where it lies.¹  The Institute for Creation Research's 2023 model, by Jeffrey Tomkins and Timothy Clarey, puts the line near the top of the Cenozoic instead.  On their account the rocks of "the majority of the Cenozoic Erathem (prior to the Pleistocene)" are late-Flood runoff, and "Tertiary mammal fossils came from creatures living at higher, more temperate elevations than dinosaurs and thus would have been buried in the uppermost Flood layers."  Horses are on their list.³³  They criticise the low boundary because it forces its holders to claim that whales "evolved rapidly from ancestors that walked out of the Ark."³  Their own model has the mirror-image problem.  If the horse beds are Flood deposits, then Eohippus, Mesohippus, Merychippus and Pliohippus lived side by side before the Flood and were buried in that order, by a sorting that no one has yet shown would follow toe count and tooth height.  Marcus Ross tested the high boundary against 303 North American mammal genera, found that 70 of them (23%) would have to cross it, and concluded that the boundary "should be located below the Pliocene/Pleistocene boundary."  In his count, 1 of 27 horse genera crosses it.¹⁵¹⁵

So each young-earth reading of the horse record has a cost.  The low-boundary reading must explain very fast change in anatomy.  The high-boundary reading must explain why a one-year Flood buried one kind's varieties in a neat anatomical order.  Neither has yet done so with data.

And not every young-earth writer accepts the series.  Jonathan Sarfati's 1999 article for Creation magazine calls the horse series "an interpretation of the data," doubts that Hyracotherium was the ancestral horse, and explains the rest as "variation within the equine (horse) kind."  He mentions in a note that "some creationists believe that there really is a trend," citing Paul Garner.¹³  He does not call the record a fraud.  On Hyracotherium the 2002 revision partly agrees with him:  the English type species is no longer filed as a horse (Part 3).  On the trend, Wise, Cavanaugh and Wood disagree with him.

What this article rests on

Every quotation printed below was checked against a source that was actually opened.  Several famous ones could not be, and they are named and left out rather than repeated.¹⁶  The two Simpson quotations that creationist writing on horses leans on most were once on that list.  Both have since been read in Simpson's own book, and Part 1 prints them side by side.

Radiometric dating is not used as evidence anywhere in this article.  That is a constraint across this whole series:  ages in years belong to the dating articles.  Every claim here works from relative order:  what sits above what in one section, and which fossils match between distant sections.  Young-earth geologists accept and use both.  Where an absolute figure appears, it is the mainstream number under discussion, not something a reader is being asked to accept.

The counts are original.  Every number in Part 2 came from a live query against the Paleobiology Database on 3 August 2026, and the query form is printed with it.¹⁷  Nobody in this argument, on either side, appears to have run them.

Part 1 — The ladder, and who took it apart

Where the picture came from

The horse diagram has a specific author and a specific date.  William Diller Matthew drew it at the American Museum of Natural History and published it in 1903.  It reached the general public through the museum's own exhibit and through school textbooks — including George Hunter's A Civic Biology of 1914, the book at the centre of the Scopes trial, which prints Matthew's diagram on page 193.¹⁸

The picture is a column of forelimbs and teeth with a small four-toed animal at the bottom and a modern horse at the top.  It reads as a staircase.  That is what it was drawn to do.

The demolition came from inside

Simpson, 1951.  George Gaylord Simpson wrote the standard popular book on the subject, Horses, and used it to attack the staircase reading.  His replacement diagram is a branching tree, not a line.  He argued that the changes were not steady, not universal, and not aimed at anything.

Two Simpson sentences circulate constantly in creationist writing on horses, and both were read in the 1951 first edition for this article.¹⁹  They belong together, because each is half of Simpson's view.

The first is the one creationists quote:

"Regarding the feet, the old idea of a steady, uninterrupted reduction in number of toes is, as has lately become clear, not only over-simplified but also essentially false.  Eohippus had four functional toes on the front feet.  This number was reduced to three and finally to one in some later horses.  So much is true, but the reduction was not universal, or constant ..."¹⁹

The second is the one they rarely quote, from the same book's introduction:

"Many other evolutionary sequences are now known, but the beautiful series of ancient and modern horses displayed in many museums are still the simplest way to convince any open-minded person that evolution is a fact.  You can see it with your own eyes."¹⁹

On the same page he writes that "no real student of horse history now doubts this for a moment."¹⁹  So the man who called steady toe reduction "essentially false" also called the museum series the simplest proof of evolution, in the same book.  He was attacking the steadiness of the change, not the change.  Notice also what he grants in the toe passage itself:  four toes, then three, then one, "so much is true."  A creationist who quotes "essentially false" and stops has cut the sentence that follows.

Prothero and Shubin, 1989.  This is the technical demolition and it is Part 4.

Gould, 1987 and 1991 and 2002.  Stephen Jay Gould wrote an essay called "Life's Little Joke," published in Natural History in April 1987 and reprinted in Bully for Brontosaurus in 1991.²  He restated the whole argument at length in The Structure of Evolutionary Theory in 2002, and that restatement was read directly for this article.²⁰

His argument has four moves, and creationists reliably quote the first and stop.

One:  the ladder is an artefact of how paleontologists pick their examples.  Gould's point is that the discipline goes looking for trends in exactly the wrong places.  A clade that has been reduced to a single surviving lineage is easy to draw as a line, because there is only one path left to trace.  A clade that is still diverse cannot be drawn that way at all.  So the groups that make the best "illustrations of progress" are the groups that have almost died out.²⁰  Horses are the standard example because there are six or eight species left out of hundreds.

Two:  the icon is worse than merely inaccurate.  This is the sentence everyone quotes, and here it is with the part that gets cut:

The ladder diagram "is much more than merely wrong.  It never could provide the promised illustration of evolution progressive and triumphant — for it could only be applied to unsuccessful lineages.  Bushes represent the proper topology of evolution."²

The clause after the dash and the sentence after it are Gould's actual conclusion.  He is not saying the fossils fail.  He is saying the diagram fails, and that the bush is the right picture — which he regards as better evidence, not worse.  Bergman's 2024 paper quotes the passage accurately and stops one clause short of the point.²¹

Elsewhere in the same book Gould gives the positive version in six words:  each stage is "a bush of several related species, not a rung on a ladder of progress."²

Three:  Gould explicitly does not deny the trends.  In The Structure of Evolutionary Theory he says plainly that he does not deny a bias toward larger size arising at speciation, and does not doubt the standard adaptive account of browsing giving way to grazing, with toe reduction and taller teeth following.²⁰  Any creationist use of Gould on horses that omits this is context-stripping.  He is arguing about the shape of the tree, not about whether the changes happened.

Four:  the counterfactual that makes the point.  Suppose Nannippus had survived instead of Equus.  Gould notes that Nannippus did not exceed the earliest horses in size and "still grew three toes on each foot," yet lasted eight million years and had molars as high-crowned as a modern horse's.²⁰  Draw the ladder to Nannippus and the same fossils produce no trend at all.  The trend is in the choice of endpoint.

Why this matters for how the article argues

A creationist does not need to establish that the ladder is wrong.  That was settled inside paleontology.  What a creationist needs to be careful about is the next step, because the demolition changes the shape of the tree and does not remove the transitions along it.  Gould, Simpson and Prothero all replaced the ladder with a bush.  A bush is still common descent.  Sliding from "the diagram is wrong" to "the sequence is worthless" is the standard failure in this literature, and Part 7 catches it happening.

Part 2 — What the record actually contains

Here is the part of the argument nobody in this debate seems to have run.  The numbers below come from live queries against the Paleobiology Database on 3 August 2026.¹⁷

Three things fall out of that table.

The record is dense enough to have been over-split.  440 species names have been proposed inside Equidae.  The database accepts 288.  Roughly a third of the names ever coined for fossil horses have been sunk as synonyms of other names.  You cannot over-split a record you do not have.  Splitting happens when there is enough material for specialists to disagree about how finely to cut it.  This is the single best number for the point that the equid record is one of the best-sampled in the terrestrial fossil record, and it is a number that cuts against our own side's usual framing.

The museum ladder shows under a fifth of it.  The classic row uses about nine genera:  Hyracotherium or Eohippus, Orohippus, Epihippus, Mesohippus, Miohippus, Parahippus, Merychippus, Pliohippus, Equus.  That is roughly 18% of the 51 accepted genera.

And the ones left out are not the marginal ones.  The three-toed hipparionine horses — Hipparion, Nannippus, Neohipparion, Cormohipparion, Hippotherium, Eurygnathohippus, Pseudhipparion — are missing from the display entirely, and they are better represented in the record than most of the animals on it.  Hipparion alone has more occurrence records than Orohippus, Epihippus, Mesohippus, Miohippus, Parahippus and Merychippus combined — 660 against 526.

That is the honest form of the "the museum is misleading you" complaint.  Not that the fossils are fake, and not that the order is invented, but that the exhibit shows under a fifth of the animals, picked to make a line, and omits a whole radiation that would ruin the line by being both abundant and off to one side.

The fifty-one genera

Ranges below are the bounds of the containing stratigraphic interval, not measured ages.  The database reports the base of the oldest interval a genus is recorded from, so a genus first recorded anywhere in the Miocene shows 23.04.  Do not read these as dates.  They are useful for ordering, which is all this series uses.

A caveat that runs against the table, printed here rather than buried.  Five accepted genera have zero occurrence records in the database.  That is a data-entry gap, not a claim that those genera are unfounded — Minippus, for instance, is one of Froehlich's own 2002 new genera, and the database has entered the name without entering its specimens.  Any diversity curve built from these occurrence counts will undercount slightly.

A second caveat for anybody re-running the query.  The database's taxon_status=valid setting does not mean "not a synonym."  Junior synonyms come back in a "valid" list carrying the senior name's occurrence count, so summing occurrences over a "valid" genus list gives 28,012 against a true total of 6,517 — a fourfold over-count.  Use taxon_status=accepted.

Reading the table

Look down the first-appearance column and the shape of the record is obvious.  The small early forms cluster at the bottom.  The Anchitheriinae come in through the Oligocene.  Then, at the Miocene, a wall:  twenty genera whose earliest record falls somewhere in an interval the database cannot resolve more finely than "Miocene."  That is not a burst of creation.  It is the resolution limit of the interval codes, and it is exactly the kind of artefact a careless argument would turn into evidence.

What the table does show cleanly is overlap.  Miohippus runs from 41 to 16 million years on these bounds.  Mesohippus runs from 39.7 to 29.5.  They overlap for ten million years on the mainstream scale.  On a ladder, Mesohippus becomes Miohippus and stops existing.  In the record, they are neighbours.  Part 4 is about what that does to the diagram.

Part 3 — Eohippus, Hyracotherium, and the name nobody gets right

This is the cleanest example in the whole topic of both sides being wrong about the same fact, in opposite directions.

Start with the fact that is not in dispute.  The Paleobiology Database does not classify Hyracotherium as a horse.  Query it today and it comes back with parent taxon Palaeotheriidae, 203 occurrences, Owen 1841 as the primary reference.⁴  The database that aggregates the published literature has followed a 2002 revision and taken the "dawn horse" out of Equidae.

Now the history, in order, because both sides garble it.  Every step below is stated in Froehlich's 2002 introduction with its original citation, and that introduction was read.²²

1. 1841 — Richard Owen names Hyracotherium, on English material. He compared it to the hyrax, which is what the name means: hyrax-like beast. It was not proposed as a horse.
2. 1872–1884 — Edward Drinker Cope describes ten North American species and files them all in Hyracotherium.
3. 1876 — Othniel Charles Marsh erects Eohippus for two North American early Eocene species. He is naming new American material. He is not renaming Owen's animal.
4. 1896 — Wortman goes back to Hyracotherium for the North American animals.
5. 1908 — Granger's revision uses Eohippus instead, for eleven species.
6. 1896, 1932 and 1952 — the synonymy is argued three times, by Earle, by Forster-Cooper, and by Simpson. All three conclude the North American and European material belong to one genus. Under the rules of zoological nomenclature, the older name wins. Owen's Hyracotherium of 1841 has priority over Marsh's Eohippus of 1876. That is why Eohippus disappeared from the technical literature.
7. 1989 and 1994 — Jerry Hooker challenges the monophyly of "Hyracotherium," and Prothero and Schoch question the synonymy in 1989. Both are saying the same thing: the genus has become a wastebasket holding several different animals.
8. 2002 — Froehlich splits it. Forty taxa, 121 characters. Hyracotherium is restricted to its type species H. leporinum, and identified as a primitive equoid rather than an equid.²² Sifrhippus is erected for H. sandrae. Minippus and Arenahippus take other pieces of the wastebasket. And — the part that matters most here — "Eohippus Marsh, the next more-derived equid taxon, is resurrected for E. angustidens Cope."²²
9. Today the database places Hyracotherium in Palaeotheriidae, outside Equidae.⁴

What the creationist version gets wrong

Bergman's 2024 paper states that "Yale paleontology professor Othniel Charles Marsh renamed the hyrax Eohippus," and that "Hyracotherium replaced the term Eohippus when the evidence convinced many paleontologists that it was not the progenitor of modern horses, nor a horse ancestor, or even a horse."²¹

Both halves are false.

Marsh did not rename the hyrax, and he did not rename Owen's fossil.  He named new North American specimens.  The hyrax is a living animal in an entirely different order and nobody has ever renamed it.

And Eohippus was not dropped because anyone decided the animal was not a horse.  It was dropped on priority — a bookkeeping rule about which of two names for the same genus was published first.  The demotion Bergman is describing is real, but it happened fifty years after Simpson settled the synonymy, in 2002, under Froehlich — and it restored Eohippus instead of retiring it.  The creationist telling has the cause backwards and the direction backwards at the same time.

What the mainstream version gets wrong

Museum labels, textbooks and popular accounts still routinely put "Hyracotherium, the dawn horse" at the base of the horse family.  On the current revision that genus is not in the horse family at all.  Froehlich's paper is twenty-four years old and the correction has not propagated to the public-facing material.

So Bergman's underlying point is correct.  The animal at the bottom of the icon is no longer regarded as a horse.  He reached the right conclusion through a false account of how it happened, and the mainstream is still printing a label that its own literature retired.

That is the honest way to write this paragraph:  concede the point and fix the history in the same breath.

One trap for anyone reading Gould on this

In The Structure of Evolutionary Theory, published the same year as Froehlich's revision, Gould asks the reader to imagine Martian paleontologists looking at two barely distinguishable species of "Hyracotherium," one of which becomes the horses and the other the rhinoceroses.²⁰  Read against Froehlich, that illustration is using the wastebasket genus in exactly the way Froehlich took apart.  It is also an argument about the Cambrian explosion, not about equid ancestry.  It should not be quoted as a statement about where horses came from.

Part 4 — Stasis, not gradualism

The horse series is offered everywhere as the classic case of gradual change.  Its own specialists found the opposite, and this section does not need a creationist source anywhere in it.

The 1989 result

Donald Prothero and Neil Shubin studied the Oligocene horses — the Mesohippus to Miohippus transition, which is the middle of the classic ladder.²³  Their conclusion, quoted verbatim by Gould in 1991 and again in 2002:

"Throughout the history of horses, the species are well-marked and static over millions of years.  At high resolution, the gradualistic picture of horse evolution becomes a complex bush of overlapping, closely related species."²³

A note on chain of custody, because this series is strict about it.  That passage was verified in two different Gould books published eleven years apart, word for word.  The original chapter itself was not opened for this article.²³  Two independent printings agreeing exactly is strong, but it is not the same as reading the source, and this article says which one it did.

What Gould reports from that chapter, in his own summary:  the Mesohippus-to-Miohippus transition shows stasis in all species of both lines, transition by rapid branching rather than by one form gradually turning into another, and stratigraphic overlap of both genera.²⁰

And the specific finding that kills the ladder:

One set of beds in Wyoming has yielded three species of Mesohippus and two species of Miohippus, all living at the same time.²⁰

Why that is fatal to the diagram and not to common descent

Both halves have to be said, and creationist writing usually says only the first.

The half that is fatal to the diagram.  A ladder requires each rung to be replaced by the next.  If Mesohippus became Miohippus, then after the transition there should be Miohippus and no Mesohippus.  Instead there are five species of both genera in one bed.  The "ancestor" is living beside the "descendant."  A staircase cannot be drawn through that, and every museum exhibit that draws one is misrepresenting the record.

The half that is not fatal to common descent.  Branching lineages are expected to overlap.  When a population splits, the parent species does not vanish.  Living examples are everywhere:  wolves did not go extinct when dogs appeared.  So overlap refutes the picture of a single line marching forward, and says nothing at all about whether the animals are related.

Anyone using this argument has to state both halves or he is not arguing honestly.

And the direction the stasis result actually points

There is a second thing in Prothero and Shubin worth noticing, and it does not favour the mainstream telling either.  Their finding is that change is concentrated at branching events and absent in between.  Species appear, stay put for millions of years, and are replaced.  That is punctuated equilibrium, applied to the sequence most often used to illustrate the opposite.

But a caution on a related creationist claim.  It is sometimes said that punctuated equilibrium was invented to explain away the horse data.  That is false as a matter of chronology.  Eldredge and Gould published the model in 1972, out of work on Devonian trilobites and Bermudian land snails.²⁴  Prothero and Shubin's horse work came seventeen years later and applied the model.  It did not motivate it.

One check left undone, and flagged as undone.  Gene Hunt's 2007 study tested directional change against random walks against stasis across 251 fossil sequences, and it is on disk for this project.²⁵  Its equid entries specifically have not been pulled.  That would be a better test than quoting anybody's summary, and this article does not claim to have run it.

Part 5 — Checking our own side: the Ashfall claim

This is the part of the article that exists because of what checking produced.

The claim

Bergman's 2024 paper states that Pliohippus, a one-toed horse, "was found buried together with Merychippus," a three-toed horse — the point being that a supposed ancestor and its supposed descendant were entombed side by side.²¹  He cites Michael Voorhies's 1981 National Geographic article on the Nebraska ashfall site.

The site is real and spectacular.  Ashfall Fossil Beds in Antelope County, Nebraska, is a waterhole where a volcanic ash fall from what is now Idaho suffocated a whole local fauna and buried the bodies where they fell.  The Paleobiology Database has it as the Poison Ivy Quarry, collection number 18324, Cap Rock Member of the Ash Hollow Formation, with 39 recorded occurrences and a field note recording more than twenty horse skeletons.²⁶

What the database actually contains

The collection was queried directly for this article.  It yields exactly five equids:

There is no Merychippus in the collection.²⁶

So the claim as stated is false.  Bergman also misspells Voorhies's name and gives the National Geographic volume as 59 rather than 159.

And the real fact is better than the false one

Look at the table again.  Two one-toed genera are entombed in a single ash fall with three three-toed genera.  All five died in the same afternoon, at the same waterhole, under the same ash cloud.

That is a better illustration of the point Bergman was reaching for than the claim he actually made.  He wanted a one-toed horse buried with a three-toed horse.  The site gives him five horses, two one-toed and three three-toed, in one bed, and the identification does not rest on anybody's phylogeny — it rests on a quarry map.

The true fact was available in a free public database, and the false one was easier to look up in a magazine article.  That is the whole lesson of this series in one example.  Checking your own side's claims is not a concession; it is where the better argument was sitting.

And the honest limit on it.  Contemporaneity at one Nebraska waterhole does not by itself show anything is wrong with equid phylogeny.  Nobody currently claims Pliohippus descended from Cormohipparion.  The three-toed hipparionines are a side radiation, and the fact that they lived alongside one-toed horses is standard textbook material, not a scandal.  What Ashfall shows, cleanly and without any dating argument, is that the tidy line of the museum case is not what the ground contains.

Part 6 — The trends, and their reversals

Three trends run along every museum ladder:  bigger, fewer toes, taller teeth.  All three are real as net changes.  All three have documented reversals or complications — and every one of those was found by mainstream researchers.

Size: about a fifth of the lineages run backwards

Bruce MacFadden's 1986 study is the standard measurement.  He examined body size across the equid record and found that 5 of 24 species lineages are characterised by dwarfism — they get smaller — and that the pattern over 57 million years is roughly 32 million years of relative stasis followed by about 25 million years of diversification and size change that is progressive but not necessarily gradual.²⁷  MacFadden's own word for the traditional reading was "oversimplified."

He restated the result in his own words in his monograph Fossil Horses:  "Of the 24 ancestral–descendant species pairs analyzed by MacFadden (1987a), 19 showed body-size increases;  however, the remaining 5 lineages were characterized by dwarfing."  And:  "although dwarfism is the exception to the rule of phyletic (also termed phylogenetic) size increase, this phenomenon occurred numerous times within the Equidae."²⁸²⁸

Gould reports the same result — "5 of 24, or more than 20 percent, leading towards decreased size."²⁰

The citations of that number are a tangle, and one of them is plainly wrong.  The paper is Paleobiology volume 12, pages 355–369, in the issue dated 1986.  MacFadden's own book dates it 1987, and reprints his body-size figure from a review he published in 1988.²⁸  Gould cites "MacFadden (1988)," which matches no entry in Gould's own reference list but may mean that review.²⁰  Bergman cites "MacFadden, 2016," with the Paleobiology volume and pages in his reference list;  no MacFadden paper of 2016 has those pages.²¹  The number, 5 of 24, is the same in every version.

The best reversal is at the very bottom of the ladder

In 2012 Ross Secord and colleagues published a high-resolution record of the earliest horse, Sifrhippus, across the Paleocene–Eocene Thermal Maximum — a sharp global warming event.  They used oxygen isotopes from the teeth of animals found with it as a temperature proxy.  Science's own editorial summary of the paper reports "a clear decrease in equid body size … during 130,000 years of warming, followed by a distinct increase as the climate cooled at the end of the period."⁵

The first horse in the sequence gets smaller, through the whole warming interval, at the exact point where the ladder is supposed to start climbing.

A restraint that has to go with it.  The widely repeated figures — shrank 30%, from about twelve pounds to eight and a half, then back up to fifteen — come from university press releases, not from a source this article opened.  The paper itself is paywalled and could not be obtained.⁵  The direction of the change is publisher-supplied text and is safe.  The pounds are not printed here.

Toes: reduction is real, "steady reduction" is not, and the anatomy is under active dispute

The endpoints are not in question.  Early equids have four toes on the front foot and three behind.  Modern horses stand on one.

What is in question is how that happened and how complete it is — and there is a live argument about it in the open literature right now, in one journal, five years apart.

Solounias and colleagues, 2018.  They argue that digit reduction is less extreme than textbooks say.  Digits II and IV persist as the splint bones, which has been known for a century.  Their new proposal is that components of all five digits persist:  I and V as the hoof cartilages and the "wings" of the last toe bone, and the lower parts of II and IV as the two halves of the frog — the soft triangular structure in the sole of the hoof.  They also report that Mesohippus "exhibits a lesser degree of digit reduction than previously thought," with fusion lines still visible on its side-digit bones.²⁹

Vincelette, 2023.  A direct rebuttal in the same journal.  He argues the side digits of three-toed horses were load-bearing and functional rather than parts waiting to be redeployed, questions whether "hoof wings" could produce the impressions in fossil trackways that Solounias's team relies on, and — the interesting part — presents "ichnological evidence that a frog was present in tridactyl equids."³⁰  If the frog was already there in three-toed horses, it cannot be made of leftover toes.

What Bergman does with this.  He cites Solounias and colleagues to argue the toes "were not lost, but are simply vestigial," which "supports the degeneration theory."²¹

That is not what Solounias and colleagues claim.  They are arguing about how single-hoofed feet were achieved within an evolutionary account, and the same paper describes the change as adaptive for trotting on grassland.  Bergman takes one side of a live 2018–2023 dispute, reports it as settled, and reverses its conclusion.  This is precisely the kind of thing this series exists to catch, and it is on our side of the table.

Teeth: what can be said and what cannot

The adaptive story is standard.  Grasses spread, grass carries abrasive silica, teeth wear faster, taller-crowned teeth are favoured.  Gould accepted that account explicitly.²⁰

The leading specialist calls that account too simple.  MacFadden, in Fossil Horses:  "careful examination of the timing and geographic distribution of acquisition of high-crowned teeth in a diversity of mammalian clades suggests that other factors were almost certainly involved."  He notes that some mammals living beside the Eocene horses had tall teeth before grasses spread on their continent, and that the change to tall teeth in horses coincides with a change to coarser sediment full of volcanic particles, whose grit wears teeth down.  His conclusion:  "the classic notion of hypsodonty in horses as purely an adaptive response to the spread of grasslands is too simple."²⁸  That cuts against the textbook story, not against evolution.  His replacement is still an adaptation, to grit as well as to grass.

There is a further objection that the timing does not work — that high-crowned teeth appear millions of years after grasses spread.  This article does not assert a specific lag, because it could not be sourced to a paper that was opened.  The two standard references, Mihlbachler and colleagues in 2011 and Caroline Strömberg's phytolith review, are both paywalled and both failed every access route tried.³¹³²  All that could be recovered of the first is Science's one-line editorial summary:  "A survey of horse-teeth wear during the past 55 million years implies that evolutionary pressures were generally low."³¹  That is suggestive and it is not a citation for a specific lag.

Stating a gap is worth more than filling it with a number nobody here has checked.

And the shape of equid diversity is the opposite of a rise

The ladder communicates a climb toward the modern horse.  The measured diversity runs the other way.

Gould, citing MacFadden and colleagues in 1999:  in North America between about 8 and 15 million years ago, an average of 16 equid species lived at the same time.  As recently as 5 million years ago, at least six lived together in Florida alone.  Today there are 6 to 8 species of horse on the whole earth.²⁰

He calls the modern equids a remnant of a much larger past diversity, a group extinguished in the continent where it originated and surviving only through Old World emigrants.²⁰  Horses evolved in North America, died out there, and came back on Spanish ships.

That is the actual shape of the data, and it is the reverse of what the exhibit communicates.  A group at the end of a long decline is being used as the illustration of triumphant progress.  Which is Gould's whole point in "Life's Little Joke," and the reason the essay has that title.

Part 7 — Auditing the newest creationist treatment

The audit target is Jerry Bergman, "The Horse Evolution Icon Exposed," Creation Research Society Quarterly 61 (Fall 2024):  133–145.²¹  It is the newest technical creationist treatment of the subject, it is free to read, and it collects nearly every claim in circulation.  It was read in full for this article.

Six of its twenty main claims are sound.  The rest are listed with the reason each one fails.

Two older creationist papers were not audited and should be before anyone treats this as complete:  Molen 2009 in Journal of Creation, and Pendragon 2015 in the same journal.  Pendragon matters more, because he advances a positive creationist phylogeny — horses descending from anchitheres rather than from tapir-like hyracotheres — and a positive claim can be tested.  Nobody has tested it.³³

Part 8 — Objections, one at a time

"The horse series is a fraud."

No.  Nor do the leading creationist treatments say it is.  Sarfati's 1999 article calls the series "an interpretation of the data," which is a different claim.¹³  The fossils are real, the identifications are made by specialists who disagree with each other in print, and the order in the rock is checkable by anybody with a shovel and a map.  6,517 occurrence records is not a fraud.  What is misleading is a display choice — nine animals from fifty-one, arranged in a line.  Call the exhibit misleading, which is defensible and was first said by paleontologists.  Do not call the record fake, which is not defensible and which no creationist paleontologist claims.

"The horse series is the best evidence for evolution."

Overstated in the other direction, and it is worth saying why a creationist should resist agreeing with it.  Because the whole sequence lies inside one family, it establishes far less than the mammal-like reptiles or the fish-to-tetrapod material, both of which cross a class boundary.  The horse record is better sampled than those.  It is not stronger evidence, because the rank it spans is one nobody disputes.

"But the ladder is still in the museums and the textbooks."

This is a fair complaint and it should be made carefully.  Bergman asserts the American Museum of Natural History has not fixed its exhibit, while himself noting that the museum built a new horse exhibit in 2008.  Nobody in this argument, on either side, appears to have gone and looked at the current exhibit, or at a shelf of current secondary-school biology textbooks, and counted.  That is a small, cheap piece of research and it would let this complaint be stated as a measurement instead of a grievance recycled from 1990.  Until somebody does it, the honest form of the claim is:  the correction has been in the technical literature since 1951, and popular presentations lag it, with Hyracotherium still labelled a horse twenty-four years after that was revised.  The complaint has a better witness than any creationist.  Bruce MacFadden, who wrote the standard monograph on fossil horses, said in 1992 of the straight-line idea:  "Some paleontologists might say that the orthogenetic paradigm is dead now.  On the contrary, I would argue that unfortunately it has become more ingrained than it should be, both in modern scientific thought and in everyday society in general."²⁸²⁸

"Evolutionists admit the horse series is wrong, so the fossils prove nothing."

This is the slide the article exists to stop.  Simpson, Gould, Prothero and MacFadden all attacked the ladder, and every one of them replaced it with a branching tree, which is still common descent.  Gould said so in the sentence creationists cut:  bushes are the proper topology of evolution.  Quoting the demolition without the replacement is misrepresenting a source, and doing it to Gould specifically — who wrote out his own concession of the trends in the same chapter — is worse.

"The toes were never lost, just reduced.  That is degeneration, not evolution."

Two problems.  First, it misreads the paper it cites, which argues for digit persistence inside an evolutionary framework and calls the single hoof an adaptation for trotting.  Second, it takes one side of a dispute that another paper in the same journal rebutted in 2023.  Reporting a live scientific argument as if it were settled, in the direction you prefer, is exactly what our side objects to when it is done to us.

"Different numbers of toes and different tooth heights are just variation within the horse kind."

This is the strongest creationist position on the subject, and it is the one Wise, Wood and Cavanaugh actually hold.  It concedes the sequence, concedes the order, and locates the disagreement where it belongs — at the boundary below Equidae, not inside it.  A creationist arguing this way does not need any of the failed claims in Part 7.

But "just variation" undersells what is being conceded.  The variation runs from four toes to one, from low-crowned browsing teeth to tall grazing ones, and from a small animal to a large one, and the creationists' own statistics put the two ends as far apart as two kinds would be, joined only by the fossils between them.¹  On the reading those authors hold, all of it happened rapidly after the Flood.¹  The reply to a skeptic is not "it is only variation."  It is:  "yes, this much change happened within one kind, and here is the young-earth account of how fast it was and why."  The second half of that reply is not yet written.  See "What the concession costs," above.

"Punctuated equilibrium was invented to save evolution from the horse data."

False as chronology.  Eldredge and Gould published in 1972, from trilobites and land snails.  The horse work came in 1989.  Drop this argument.

"The sequence is only a sequence because evolutionists arranged it that way."

The arrangement is a hypothesis; the stratigraphic order is not.  Small, low-crowned, many-toed equids occur low in the section and large, high-crowned, one-toed equids occur high.  Within a basin that is superposition.  Between basins it is the matching of fossils from place to place, which is older than radiometric dating and which young-earth geologists also use.  Cavanaugh, Wood and Wise ranked their own horse species against that matched order and found it fit.¹  The companion order article on this site ran the strict test across 2,001,053 occurrence records and found two order violations in the whole database, both traceable to data-entry errors, neither of them an equid.³⁴  A young-earth account still owes an explanation of why the order is what it is, and hydrodynamic or ecological sorting has not been shown to predict this particular one.

What would change this verdict

• Opening Prothero and Shubin's 1989 chapter itself, rather than reading the key passage through two Gould printings.
• Obtaining Secord and colleagues' 2012 paper, so the Sifrhippus dwarfing can be reported with its measured magnitudes instead of its direction only.
• Obtaining Mihlbachler and colleagues' 2011 paper or Strömberg's review, which would settle whether the tooth-height timing objection is real. Until then this article does not make it.
• A young-earth rate for the horse series. A creationist who holds the low Flood boundary needs a measured or modelled rate at which toe loss and tall grinding teeth could arise in a few centuries. A good one would turn the article's weakest young-earth point into a testable claim.
• A burial account from the high-boundary model. If the horse beds are Flood deposits, as the Institute for Creation Research's model holds, its authors need to show what sorting would bury one kind's varieties in order of toe count and tooth height, on more than one continent.
• A larger baraminology run. Cavanaugh, Wood and Wise used nineteen species and a dataset that is 64% teeth, and asked for a repeat with more taxa and more of the skeleton. If the two ends still separate and only the intermediates join them, the skeptic's reading of that result gets stronger.
• Opening MacFadden's 2005 Science paper, so Bergman's citation of its page 1729 can be judged on the text.
• A count of current museum exhibits and current textbooks. It would turn the article's central complaint from an anecdote into a measurement.
• An audit of Pendragon's 2015 anchithere-origin proposal, which is the only positive, testable creationist phylogeny of horses in print.
• A serious pass on where perissodactyls come from. That is the gap creationism actually disputes, and this article does not touch it.

Sources

Research dossier, with the full genus table, the taxonomic history traced through the primary literature, nineteen quotations with verdicts, and every source that could not be reached:  research/horse-sequence.md.

Quotations deliberately left out.  Five famous passages that appear in nearly every creationist treatment of horses are not printed above, because no source containing them was opened for this article:  Simpson's 1951 line on toe reduction being "essentially false"; Simpson's 1961 line that the museum series is the simplest way to convince an open-minded person that evolution is a fact; Heribert-Nilsson's "continuous only in the textbooks," which reaches the modern literature through a 1971 creationist secondary whose reference list dates the source book two different ways; David Raup's 1979 remark about having fewer transitional examples than in Darwin's time; and Ernst Mayr's 2001 comment on gaps in phyletic series.  The Simpson 1961 line is the one this article most wanted, because it is the honest counterweight to everything else in Part 1.  It is left out anyway.  A quotation that cannot be checked is not evidence, whichever side it helps.

What was read in the original.  Froehlich 2002, Bergman 2024, Solounias et al. 2018 and Vincelette 2023 were obtained and read in full.  Gould's The Structure of Evolutionary Theory was read from a digital copy held in this project's source library.  Bully for Brontosaurus was verified passage by passage through in-volume search, which returns printed page numbers.  For MacFadden 1986, MacFadden 2005, Mihlbachler et al. 2011, Strömberg 2011 and MacFadden et al. 1999, no copy could be obtained by any route, and every footnote below says so where it applies.