Creationist Biogeography: The Post-Flood Dispersal Model

"It is not suggested here that land animals survived the Genesis Flood on rafts, but that rafts would have facilitated their dispersal after the Flood, as they multiplied and migrated away from the ark where it settled on mountains of Ararat." — Dominic Statham, stating the model this article tests¹

"Many terrestrial organisms may have rapidly migrated from the ark to the rafting 'launch' points, crossed the oceans on the rafts, and established themselves at huge distances from the ark without leaving evidence of the dispersal process in the fossil record." — Kurt Wise and Matthew Croxton, 2003²

"It's not only unlikely, it is highly improbable." — Chad Arment, a young-earth creationist, on his own movement's late-Flood-boundary models, after running the marsupial numbers³

Summary

Noah's Ark landed at one point in eastern Turkey or Armenia. Every land animal alive today is supposed to descend from the pairs that walked off it, about 4,500 years ago. So how did kangaroos get to Australia, lemurs to Madagascar, and nothing at all to New Zealand except bats?

Creationists have a worked answer. The fullest technical version is Kurt Wise and Matthew Croxton's 2003 paper for the Fifth International Conference on Creationism, which proposes that the post-Flood oceans were covered in log mats and that animals rafted across them.² The best-known popular version is Dominic Statham's 2010 article in Journal of Creation, reprinted on creation.com, which adds three more mechanisms: transport by humans after Babel, land bridges exposed by a post-Flood Ice Age, and rapid speciation from a small number of created kinds.¹

This article takes that model seriously, states it in its authors' own words, and then tests it. Two of the tests are original measurements made for this article.

The conclusions, in order of confidence

1. The land-bridge mechanism fails, and the leading creationist authority on it agrees. Measured here from ship and satellite bathymetry: to walk from the Asian mainland to Australia, sea level has to fall 1,264 m. The controlling sill is in the Banda Sea at 132.3°E, 7.4°S. The Last Glacial Maximum lowstand was 134 m,⁴ so the required drawdown is 9.4 times the deepest the Ice Age is known to have produced. Michael Oard's own post-Flood Ice Age model yields only about 50 m, which he states himself.⁵ Against that the shortfall is a factor of 25. Oard has drawn the obvious conclusion. In 2020 he listed three proposed post-Flood land bridges and dismissed the Australian one before discussing it:⁵ "The third land bridge does not seem plausible because of the many deep, wide ocean channels."⁵ In 2025 he repeated it: "the idea of a land bridge between Sunda and Sahul is unlikely."⁶⁶ This article's measurement does not refute Oard. It puts a number on a judgement he had already made, and the number is 1,264 m.

2. The first water gap on the route already defeats it. The shallowest walkable route from Bali to Lombok, measured on a 61 m grid, has a sill of 226 m. That single 70 km strait, the first step past the edge of the Sunda shelf, needs 1.7 times the LGM drawdown and 4.5 times Oard's. Everything east of it, including all of Australia, is on the far side.

3. The same measurement works perfectly where creationists say land bridges worked. Bering Strait: 44 m. Britain to Europe: 30 m. Sumatra to Borneo: 33 m. New Guinea to Australia: 11 m. Sri Lanka to India: 3 m. Tasmania to Australia: 60 m. Every one of these is a real land bridge at glacial lowstand, and every one has the shared continental fauna to prove it. The method is not rigged against the model. It confirms the model's land bridges everywhere they exist, and it finds none where the model needs them most.

3a. So the whole weight now rests on rafting, and every creationist writing on this says so. Oard lists five candidate mechanisms and, having ruled out the land bridge, keeps four: "(1) a land bridge, (2) island hopping, (3) transport on log mats, (4) human transport, or (5) God's providential direction."⁷ He rejects human transport for carnivorous marsupials and venomous snakes.⁶ What is left is log mats, with divine direction held in reserve. His 2025 verdict: "The only viable solution for the arrival of marsupials to Australia seems to be transport on log mats after the Flood."⁶

4. Rafting is a real mechanism, it has mainstream support, and the distance it has to cover is short. Statham reports a witnessed iguana landing on Anguilla after the 1995 hurricanes; the proposed origin on Guadeloupe implies a crossing of about 320 km.¹ Nearly 300 Japanese coastal species crossed the Pacific on tsunami debris and were recovered alive in North America six years later.⁸ Alan de Queiroz's 2005 review in Trends in Ecology and Evolution records mainstream biogeography swinging back toward oceanic dispersal across a wide range of groups.⁹ Anyone who calls rafting a creationist invention has not read the literature.

4a. And measured here, the water the creationist model needs an animal to cross is less than a hundred kilometres. On the easiest island-hopping route from the Sunda shelf to the Sahul shelf, the longest single open-water leg is 85 km at Oard's 50 m of drawdown, 78 km at the glacial lowstand, and 110 km at today's sea level. That is shorter than the 320 km the Anguilla iguanas were observed to cover, and about seventeen times shorter than the roughly 1,450 km that mainstream biogeography needs for its Atlantic crossings.¹⁰¹¹ The water gap is not this model's problem. Section 9 works through what that does and does not settle.

5. But the rafting model's own arithmetic gives it a short window. Wise and Croxton take their flotation half-life from Steven Austin's Spirit Lake measurements: 75 years for Douglas fir.² Run that forward. At 100 years after the Flood, 40% of the mat is still afloat. At 500 years, 1%. At 1,000 years, one part in ten thousand. The mechanism is gone long before recorded history begins, which is convenient, and it is also gone before the Ice Age peaks in Oard's own timeline, which is not.

6. The dispersal trail is missing, the creationists concede it, and the size of the gap can now be stated as a number. Measured here from the Paleobiology Database: Australia has 1,219 crown-marsupial occurrences dated inside the Pleistocene. All of Asia, across every epoch from the Cretaceous to today, has ten, the youngest a single Miocene tooth genus in Thailand. Between Turkey and Australia there is nothing in the Pliocene, nothing in the Pleistocene, nothing in the Holocene. Oard states the problem in his own words: creationists "must explain how the marsupials left the Ark in the Middle East and arrived in Australia with little or no sign of them in between."⁶ His answer is the one that cannot be tested: "a rapid migration after the Flood may have left no fossils. After all, God was possibly directing their dispersal."⁷

7. Marsupial genera are 94% single-continent, which is high but not unique. Measured here: 261 of 278 crown-marsupial genera in the Paleobiology Database occur on exactly one continent. That sounds decisive until it is calibrated. Primates score 91%, xenarthrans 89%. Some of this is a sampling artifact, because a genus known from one quarry is single-continent by default. The signal is in the contrast: Carnivora scores 68% and Proboscidea 60%. Groups that actually crossed land bridges left a multi-continent fossil record. Marsupials did not.

8. A creationist got there first, and his arithmetic is right in direction and wrong in method. Chad Arment's 2020 paper in Answers Research Journal counts marsupial genera that appear on the same continent both below and above a proposed Flood boundary, and calculates (1/6)^x for x genera.³ His conclusion, that a Cenozoic Flood boundary is untenable, survives an independent recount from the raw database. His method does not. The (1/6)^x formula assumes each genus dispersed independently, and animals do not disperse independently.

9. Two quotations that carry a lot of weight in this literature do not survive a check of their context. Both are verbatim and both are on the pages cited. Neither says what it is used to say. Details in the quote audit below.

10. Overall verdict: the model is a real model, it makes real predictions, and on the two predictions this article could test with primary data it fails. That is not a reason to abandon the doctrine of a global Flood, which rests on other evidence. It is a reason to stop offering post-Flood biogeography as a solved problem. The honest position is the one Wise and Croxton themselves took in 2003 when they published eighteen tests and asked the creation-biology community to run them. Twenty-three years later almost none of them have been run.

Every number labelled "measured here" was computed for this article. The scripts are in research/ beside it, they take no hidden inputs, and they run in minutes.

1. What the model actually says

The model is easy to caricature and worth stating carefully.

1.1 The problem it is solving

Genesis 8:4 puts the Ark on the mountains of Ararat. Genesis 8:19 has every animal leaving it. If that is history, then every terrestrial vertebrate lineage on Earth radiates from one point in eastern Anatolia within the last five thousand years.

The pattern that has to be produced is not vague. Marsupials live in Australasia and the Americas and nowhere else. All hundred-odd living lemur species are confined to Madagascar. New Zealand had no native land mammals at all except bats. Hawaii had one bat and one seal. Around a third of the world's described Drosophila species are Hawaiian endemics.¹ Each continent has a fauna that is recognisably its own.

Creationists know this is the hard case. Wise and Croxton open by conceding that earlier creationist models could not do it: the fixity-plus-dispersal model of Linnaeus's era "was not challenged by pre-Columbian biogeographic knowledge" and collapsed as soon as the New World was explored, because similar environments on opposite sides of the Atlantic turned out to hold entirely different animals.²

1.2 The four mechanisms

Statham lists four processes.¹

Transoceanic transport on vegetation mats. The Flood tore up the world's forests. Billions of trees were left floating. Ocean currents moved these mats around the globe, and animals boarded and left them at the coasts. This is the load-bearing mechanism.

Transport by man. After the dispersion at Babel, people spread over the whole earth and took animals with them, as food for the journey and stock for arrival.

Migration and partial extinction. A post-Flood Ice Age lowered sea level and opened land bridges. Animals crossed, then went extinct in the places between, leaving today's disjunct distributions.

Speciation. Only kinds were on the Ark, not species. The 250-odd Australian marsupial species need not all have been aboard. They diversified after landing.

1.3 Wise and Croxton's version, which is the serious one

Statham's article is a summary for a general readership. Wise and Croxton's paper is a research programme, and it deserves to be judged as one.

They call their model "diversification/rafting" and set it beside five historical alternatives in a table scoring each against eight classes of biogeographic evidence.² Their claim is that theirs is the only model that scores "yes" on all eight, and specifically the only one that explains three things: why areas of endemism exist where they do, why widely separated areas of endemism resemble each other, and why so many groups have ranges broken across oceans.

The mechanism is quantified, after a fashion. Around a million logs floated on Spirit Lake after the 1980 Mount St. Helens eruption. Twenty-three years later hundreds of thousands were still afloat. Steven Austin, in a personal communication they cite, fits an exponential decay to the mat and estimates a flotation half-life of 75 years for Douglas fir.² Scale that to a global forest and, in their words, "log mats immediately after the Flood may have been nearly as efficient for dispersal of some terrestrial organisms as was the land itself."²

Then they do something almost no creationist paper does. They list eighteen predictions and invite falsification. A sample:

• Areas of endemism should sit where ocean currents run into land.
• The frequency of transoceanic disjunction should track how raftable an organism is.
• Direction of dispersal, read off phylogenies and fossils, should match current direction.
• Rafting success should fall off exponentially through the Cenozoic.
• The proportion of groups with broken ranges should fall from families to genera to species.
• Australian marsupial fossils should be confined to post-Flood strata and to Australia.

That last one is a prediction about the fossil record that a database query can now check, and section 5 checks it.

They also concede two things that matter later. God chose which individuals boarded the Ark, and may have chosen ones "pre-designed for the post-Flood dispersal event."² And animals may have crossed oceans "without leaving evidence of the dispersal process in the fossil record."² Both are unfalsifiable as stated. A model is entitled to some of that. A model that leans on it at the exact points where it is being tested is in trouble.

2. The problems the model is answering are real problems

It would be easy, and dishonest, to skip this section. Several of the observations creationists cite against conventional biogeography are genuine puzzles that the field has argued about for a century.

Areas of endemism. In 1820 Augustin de Candolle identified twenty regions of the globe where a large share of the plant species are found nowhere else.² Those regions later turned out to match areas of high animal endemism too. A simple "species originate at random across a continent" model does not predict twenty concentrated hotspots.

Disjunct lineages younger than the continental split. Statham's example is cacti: South America and Africa separated around 100 million years ago, and cacti are dated to about thirty million years, yet cactus species occur in Africa.¹ The general form of the complaint is sound, and mainstream biogeography now agrees with it. Alan de Queiroz's 2005 review is built on exactly this observation, and his verdict is blunt: molecular dating "favors oceanic dispersal over tectonic vicariance as an explanation for disjunct distributions in a wide variety of taxa, from frogs to beetles to baobab trees."⁹ For baobabs, the Africa-Australia split is dated at 5 to 23 million years against a tectonic separation 120 million years ago. Even the oldest molecular estimate is off by a factor of five.⁹

Trans-Pacific tracks. Léon Croizat spent decades drawing lines connecting the same disjunct patterns across group after group, and identified five recurring nodes or "gates."² Croizat was a serious and difficult figure whose work most biogeographers rejected, and the patterns he mapped are still real patterns.

The marsupial geographical switch.  Statham contrasts the living southern distribution with northern Late Cretaceous metatherian fossils, and quotes Cifelli and Davis's 2003 phrase "this geographical switch remains unexplained."¹  The quotation is their historical assessment, as reported by Statham.

Section 5 puts numbers on it, and the numbers are stark: 773 Cretaceous metatherian occurrences in North America, 28 in Asia, 5 in Europe, zero in South America, Australia or Africa.

The eastern Asia / eastern North America connection. Hong Qian's 2002 study in The American Naturalist counted 627 seed-plant genera shared between eastern Asia and eastern North America, 151 of which do not occur in western North America.¹² Some of the shared taxa are identical at species level. Statham argues this implies a recent split, because plants change fast.¹ The inference is too quick, since many plant species are morphologically static for long periods, but the pattern itself is odd and has been debated since Asa Gray raised it in 1846.

None of this is manufactured. A young-earth model that engages these problems is engaging with real ones.

3. The land-bridge mechanism, measured

This is where the model breaks, and the break can be measured rather than asserted.

3.1 The right question, and how to answer it

Arguments about the Wallace Line usually trade quotations about how deep the Lombok Strait is. That is the wrong question. A strait's average depth does not matter, and neither does its depth at the narrowest point. What matters is the sill: over every possible route from A to B, the one whose deepest point is shallowest. Drop sea level by that amount and there is a continuous walkable path. Drop it by any less and there is not.

That is a standard problem in computer science, the minimax path, and Dijkstra's algorithm solves it exactly if you replace addition with max(). Run it on a bathymetry grid and the answer is unambiguous.

Bathymetry here comes from the GMRT Grid Server, which merges ship multibeam soundings with satellite-derived gravity data. Regional runs use a grid of about 7.8 km per cell. The Bali-Lombok run uses about 61 m per cell, because coarse grids smooth narrow channels and make them look shallower than they are. The coarse grid put the Bali-Lombok sill at 290 m; the fine grid found a genuinely shallower saddle at 226 m. Where they disagree the fine grid wins, and this article always quotes the shallower, more model-friendly number.

3.2 The results

Measured here.¹³ The 134 m lowstand is Lambeck and colleagues' 2014 figure in PNAS, taken from their own text: the eustatic sea level "reaches its lowest value of ~134 m" at the end of the interval 29 to 21 thousand years ago.⁴

3.3 What the table says

Read the top half first. Every crossing that creationist authors cite as a land bridge is a land bridge. Beringia, the Sunda shelf, the Sahul shelf, Doggerland, the Palk Strait, Bass Strait: all of them emerge with room to spare at a 134 m drawdown, and several emerge at a third of it. The method is not built to embarrass the model. It confirms the model wherever the model is right.

Now read the bottom half. The Sunda-Sahul sill is 1,264 m. There is no route, anywhere in the Indo-Australian archipelago, that lets a kangaroo's ancestor walk from Asia to Australia without a 1,264 m fall in sea level. The controlling point is in the Banda Sea at 132.26°E, 7.39°S.

How much ice is that? Lambeck's 134 m of drawdown corresponds to 52 million cubic kilometres of extra grounded ice.⁴ Scaling linearly, 1,264 m requires about 490 million cubic kilometres. Spread that over every square kilometre of land on Earth, tropics included, and it averages 3.3 km thick. Nobody proposes this. It is not a young-earth problem or an old-earth problem. It is a problem for anyone who wants terrestrial animals to walk into Australia, and mainstream biogeography does not want that either.

3.4 The Wallace Line is the first step, not the last

The specific number that kills the model is not 1,264 m. It is 226 m.

The Sunda shelf is one landmass at a 33 m drawdown: Malaya, Sumatra, Java, Borneo and Bali all join. Then comes the Lombok Strait, about 70 km of open water at the latitude where the shallowest route crosses, with a sill at 226 m. That is where Alfred Russel Wallace drew his line in 1859. Everything west of it has Asian mammals: tigers, elephants, rhinos, orangutans. Everything east of it does not.

An animal walking out of Anatolia gets, at best, as far as Bali. Then it needs to cross water. So do all the animals behind it, and so does every subsequent step across Wallacea, and the sills only get deeper. Borneo to Sulawesi is 722 m.

Michael Oard's post-Flood Ice Age model, the standard creationist source for lowered sea level, produces a fall of only about 50 m. That is not an opponent's estimate but his own, and he gives the reasoning: "the drop in sea level at glacial maximum would have been much less than secular scientists believe. It may have been only about 50 m, assuming Northern Hemisphere ice sheets were on average about 700 m thick and the Antarctic ice sheet was about 1,200 m thick at glacial maximum."⁵ Against 50 m, the Lombok Strait is short by a factor of 4.5 and the full Sunda-Sahul route by a factor of 25.

Fifty metres is not enough even for Beringia at 44 m, on Oard's own reckoning, once isostatic and tectonic complications are added. That is why his 2020 paper abandons a sea-level Bering bridge and proposes a tectonically raised one instead: "I once believed that the land bridge was exposed late in the Ice Age because of a lower sea level. However, now I think it more likely that it was raised early in the Ice Age for several reasons."⁵

3.4a Oard reached the same verdict without the arithmetic

It would be a misrepresentation to leave the impression that creationists are still offering a land bridge to Australia. The best-informed of them stopped.

Oard's 2020 survey lists three proposed post-Flood land bridges, discusses two, and disposes of the third in one sentence: "The third land bridge does not seem plausible because of the many deep, wide ocean channels, but will be discussed further in another paper on the subject of the dispersal of the Australian marsupials."⁵

He kept that promise three times over. In 2022 he wrote that "east of Wallace's Line, several deep, wide ocean channels would have halted migration," and drew from it an argument this article makes independently: "The fact that many of today's southeast Asian animals did not migrate to Australia demonstrates the limits of this late Ice Age land bridge."⁷ In 2025 he added the second half of the argument, which is Darwin's: "if there was such a land bridge, the fauna and flora would be more homogenous between Sunda and Sahul."⁶

So the measurement in section 3.2 is not a refutation of Oard. It is a number attached to a judgement he had already reached on qualitative grounds, and the number is larger than the qualitative language suggests. "Many deep, wide ocean channels" is a fair description of a 226 m first step. It understates a 1,264 m route. The value of the measurement is that it converts an impression into a quantity that can be set beside 134 m and 50 m, and it removes any remaining room to argue that some combination of ice volume and isostasy might yet close the gap.

3.5 The escape route, and why it is expensive

There is one honest reply available, and it should be stated at full strength.

Catastrophic plate tectonics holds that the ocean basins deepened rapidly during and after the Flood. If the Banda Sea sills were shallow in the first centuries after the landing and subsided afterwards, the measurement above describes a world that did not yet exist.

This cannot be ruled out, and it costs a great deal.

It has to be finely tuned. The sills must have been shallow enough to walk in the window when marsupials were crossing, and deep enough now to give a 1,264 m sill, with the subsidence timed to fall between. Nobody has modelled this.

It contradicts the creationist authors themselves. Wise and Croxton say that after the Flood, apart from India, Panama and the Tethys seaway, "modern continental configurations are fairly close approximations to continent/ocean relationships during times of post-Flood dispersal."² Their whole rafting model depends on modern ocean currents being a first-order guide to post-Flood ones, which requires modern geography.

And it makes a prediction that can be checked. Drowned land bridges leave evidence. The Sunda shelf carries drowned river valleys, terrestrial sediments and peats. The Banda Sea does not. It is oceanic crust under abyssal sediment, and nothing in its stratigraphy looks like a subaerial land surface of any age, let alone a recent one.

4. The rafting mechanism, on its own terms

Rafting deserves better than the dismissal it usually gets from critics. Then it deserves the arithmetic.

4.1 The evidence for it is good

Statham reports that fishermen on Anguilla saw iguanas arrive on logs and uprooted trees in 1995, a few weeks after two hurricanes struck the Lesser Antilles.¹  The landing was witnessed; the proposed Guadeloupe origin and roughly 320 km route were inferred.

In 2011 the Tōhoku tsunami swept Japanese docks, boats and debris into the Pacific. James Carlton and colleagues, reporting in Science in 2017, recovered 289 living Japanese coastal species on debris that reached North America and Hawaii, some of it six years later.⁸ That is the largest documented transoceanic rafting event in history and it happened while people were watching.

Mainstream biogeography relies on rafting for Madagascar. Jason Ali and Matthew Huber argued in Nature in 2010 that ocean currents, not land bridges, delivered Madagascar's mammals, and stated the point against land bridges explicitly.¹⁴ Franck Poux and colleagues had already shown that Madagascar's four endemic mammal clades arrived at four different times, which is what repeated chance crossings look like and not what a land bridge looks like.¹⁵

Anyone who tells a creationist that rafting is an ad hoc invention is wrong, and Wise and Croxton were right to build on it.

4.2 The arithmetic the model's own half-life implies

Take Wise and Croxton's 75-year flotation half-life at face value and run it forward.

The Spirit Lake row is a check on the model, not on the mat: a million logs in 1980, 81% of them still afloat 23 years later, gives about 810,000, and Wise and Croxton report "hundreds of thousands." The half-life is internally consistent.

The consequences are less comfortable. The mat is 99% gone within 500 years. In Oard's timeline the Ice Age peaks around then and the land bridges open. So the two mechanisms do not reinforce each other. They barely overlap. Rafting is nearly finished by the time the land bridges arrive, and the land bridges do not go where the animals need to go.

A second consequence is quietly favourable to the model and should be said. If rafting really did decay exponentially, then the fact that transoceanic dispersal looks like a "sweepstakes route" today, in George Gaylord Simpson's phrase, is exactly what the model predicts.² Wise and Croxton make this argument themselves, and it is a good one. It is also unfalsifiable with present data, because no independent clock on the decay exists.

4.3 What rafting delivers, and what it does not

Here the mainstream evidence turns against the model.

Carlton's tsunami rafts were overwhelmingly wooden vessels and plastic buoys, and the paper's central point is that plastic changed the game: wooden debris waterlogs and sinks within a few years, which is why 150 years of Pacific shipping never delivered a living Japanese community to North America and six years of plastic did.⁸ The post-Flood log mat is wood.

Ali and Huber's Madagascar model has a second half that creationists rarely quote. Rafting worked when the Palaeogene current system pointed the right way. Once Madagascar drifted north and the modern gyre established itself, it stopped. Their words: this "may explain why no fully non-aquatic land mammals have colonized Madagascar since the arrival of the rodents and carnivorans during the early-Miocene epoch."¹⁴ A post-Flood dispersal 4,500 years ago happens under the modern current regime, which is the regime their paper says does not work.

And rafting has a size limit. Alan de Queiroz, and the sources Arment himself cites, note that successful oceanic founders are small to medium-sized animals that later diversify into larger ones.⁹ Green iguanas ride out hurricanes. Diprotodon does not. Oard's reply is that post-Flood mats were in a different class from anything observed since, "over 100 km² and 10 m thick," and that moderate-sized animals must be able to raft because ground sloths reached the West Indies.⁶ The sloth point is a good one. The mat dimensions are an estimate with no stated derivation.

4.4 The current-direction problem, conceded

The sharpest difficulty is not size. It is direction, and Oard states it against himself in his own abstract: "Indian Ocean currents today would not carry a log mat from the shores of southwest Asia to Australia, but the currents were likely different early in the Ice Age."⁷

That is the same difficulty Ali and Huber's Madagascar paper raises, arriving from the other side. Their model works because Palaeogene currents ran the right way and stopped working when the modern gyre formed.¹⁴ Both cases need a current regime that no longer exists.

For an old-earth model this is a cost that can be paid, because the Palaeogene current field is independently reconstructible from plate positions and sediment drift deposits. For a 4,500-year model it cannot, because the continents are already in their modern positions, which Wise and Croxton state explicitly,² and a modern ocean basin with modern continents produces modern currents. Oard's escape is that the early post-Flood ocean was much warmer, which would genuinely alter circulation. Nobody has modelled how much, or in which direction, and until somebody does, "the currents were likely different" is a hope rather than a mechanism.

5. The dispersal trail: a fossil census

The most common objection to the whole model is that there is no trail. The animals should have left fossils along the way. This is usually asserted. Here it is counted.

5.1 Method

Every metatherian occurrence in the Paleobiology Database was pulled on 25 August 2026, 4,857 records in the Metatheria download; a separate Marsupialia download contains 3,564 records.  These are independent query results rather than a strictly nested pair of record sets.¹⁶ Occurrences were assigned to an epoch only when the whole PBDB age bracket fits inside that epoch, which discards about 10% of the records and cannot invent a record that is not there. Country codes were mapped to continents.

5.2 Crown marsupial occurrences, by epoch and continent

Measured here.¹⁶

5.3 What the Asian column contains

All ten of them, in full, because the whole argument turns on this cell:

Measured here.¹⁶ Two things stand out.

Every one is a herpetotheriid or a peradectid. These are northern-hemisphere stem groups, not kangaroo relatives. Arment says so plainly, and puts it better than a critic could: comparing kangaroos to Herpetotherium or Siamoperadectes "would be like comparing distinctly different placental mammals such as cats to elephants."³

And the record stops in the Miocene. In the Pliocene, Pleistocene and Holocene, the interval that contains the entire post-Flood period on any creationist chronology, Asia has zero marsupial fossils. In the same window Australia has 1,488.

5.4 The comparison that matters

1,219 Pleistocene occurrences in Australia. Zero across the whole of Asia. That is not a gap in a poorly sampled record. Asia in the Pleistocene is among the most intensively excavated ground on the planet, and it has produced a rich Pleistocene mammal record of everything except the animals that were supposed to be walking through it.

The model has an answer, and Wise and Croxton wrote it in 2003: animals may have crossed "without leaving evidence of the dispersal process in the fossil record."² Oard repeats it in 2022, and adds the second clause: "Although Ice Age marsupial fossils are not found in Southeast Asia west of Sulawesi, a rapid migration after the Flood may have left no fossils. After all, God was possibly directing their dispersal."⁷

That is a coherent thing to say once. Said at the one place where the model is being tested, it stops being a model and becomes an assurance. And it is worth noticing that Oard himself rules this move out of bounds when a rival creationist makes it: he treats appeals to divine guidance in dispersal as untestable, and Arment says so more bluntly still, calling them "simply God-of-the-gaps theorizing."³

Two qualifications that a critic would raise, and one of them has force.

First, the census is a database census, and the Paleobiology Database is not a complete inventory. It does not contain the Holocene marsupial remains from a cave on Halmahera reported by Flannery and colleagues in 1995, nor the Sulawesi cuscus, both of which Oard cites.⁶ So "zero in Asia" is a statement about PBDB, and the honest form of it is that PBDB records no marsupial anywhere between Anatolia and Wallace's Line at any date. That is still the claim the argument needs, because Halmahera and Sulawesi lie east of Wallace's Line and their marsupials are Sahul-derived elements that moved west. They are not a trail from Ararat. They are the far end of the problem.

Second, Skeels and colleagues combined data on 20,433 species in all 227 terrestrial vertebrate families in the archipelago with phylogenetic and climate models.  They reconstructed "a minimum of 381 independent colonization events across Wallace's Line" and linked the asymmetry to climatic history.¹⁷  Their introduction separately reports earlier work estimating exchange "at least twice as high from Sunda to Sahul than in the opposite direction."¹⁷  Oard repeats the 381 figure but calls the events terrestrial-mammal colonizations; Skeels's estimate covers terrestrial vertebrate families.⁶  The model supplies a quantitative comparison, conditional on its phylogenies and reconstructions.

5.5 Antarctica, which is the model's best remaining route

The census does turn up something the model can use. Antarctica has 16 crown-marsupial occurrences, all Eocene, spanning five named genera: Derorhynchus, Marambiotherium, Pauladelphys, Woodburnodon, Xenostylos.

Arment noticed this and proposed the Antarctic route: Ararat, then South America, then Antarctica, then Australia, before the continent froze.³ It has real support. Two genera, Derorhynchus and Pauladelphys, occur in both South America and Antarctica, according to this census. Woodburnodon does too. Maria Nilsson and colleagues, using retroposon insertions that are effectively immune to the homoplasy that plagues sequence data, found that all four Australasian marsupial orders share a single origin, and that the pattern indicates "a single marsupial migration from South America to Australia."¹⁸ One migration, South America to Australia, is exactly the shape of route Arment needs.

Two cautions on Nilsson's result, because it is easy to over-read and this article over-read it in an earlier draft. It is derived from living marsupials, and the current review of the field says the fossils do not agree: "It is also clear there was not simply a single dispersal of marsupials from South America to Australia (the scenario implied by phylogenies of extant marsupials only); instead, the fossil evidence from Tingamarra, South America and Antarctica suggests a much more complex biogeographical relationship between these landmasses, characterised either by multiple dispersals, differential extinctions, or a combination of these."¹⁹ Beck and colleagues go further and suggest the traffic may have run the other way, from Australia to South America.²⁰ So the single-migration result is what the molecular data from surviving lineages imply, and it is contested by the fossils. A creationist is entitled to point that out.

It is also, on a young-earth clock, a route with no time. Arment concedes it: "there would likely only be a few hundred years available between the end of the Flood and the complete isolation of Antarctica."³ In those few centuries the animals must leave Anatolia, reach South America across the Atlantic, walk the length of South America, cross into Antarctica, traverse a continent, cross into Australia, and radiate into 38 Oligocene genera.

The overland legs of that are the easy part, and the model is right about it. Introduced animals spread across continents fast. Arment's own example is the rabbit, which crossed Australia from Victoria to the west coast in about thirty-five years from thirteen founders.³ Walking is not the bottleneck. Water is. And leave, along the whole of that route, five Antarctic genera and nothing else.

6. Arment's calculation, replicated and criticised

Chad Arment's 2020 paper is the best piece of quantitative work the creationist biogeography literature has produced, and it is aimed at other creationists.

6.1 What he did

If the Flood boundary sits in the Cenozoic, then fossils below it are pre-Flood and fossils above it are post-Flood. A genus that appears on the same continent on both sides must have lived there before the Flood, boarded the Ark, and returned to the same continent afterwards. Arment charted 294 marsupial genera and counted how many do that at each candidate boundary. Then, following a suggestion by Tim Clarey at the 2018 International Conference on Creationism that any one animal has a one-in-six chance of returning to its home continent, he raised it to the power of the number of genera.³

His answers: (1/6)⁴⁶ at the Oligocene-Miocene boundary, (1/6)³¹ at Miocene-Pliocene, (1/6)⁶¹ at Pliocene-Pleistocene. The last is 3.4 × 10^-48.

6.2 Recomputed here

The same census was run from the raw database, with the conservative epoch rule described above and restricted to crown Marsupialia.

Measured here.¹⁶

The recount is lower everywhere, and the gap is explained rather than mysterious. Arment used the Fossilworks portal in 2018 plus additional published sources; this recount uses PBDB alone in 2026. He counted 294 genera of Metatheria; this counts 280 genera of crown Marsupialia. He accepted an occurrence's epoch assignment; this recount throws out every occurrence whose age bracket straddles a boundary. Every one of those choices makes the recount smaller. It is a floor, not a best estimate.

The direction of his result survives completely. At the Pliocene-Pleistocene boundary, 40 marsupial genera appear in Australia on both sides. A late Flood boundary requires all forty to have made a round trip to Ararat and back.

6.3 Where the method is wrong

The (1/6)^x formula treats each genus as an independent draw. Animals do not disperse independently. They move as communities, along the same corridors, behind the same climate. If one marsupial lineage found a route to Australia, the rest had a far better than one-in-six chance of following it. The true joint probability is enormously larger than (1/6)⁴⁰ and cannot be computed without a dispersal model nobody has.

Two things survive the criticism.

The comparison between boundaries still holds, because the same bias applies at each one. Whatever the correct per-genus probability is, forty crossings are worse than nine.

And the correlated-dispersal defence has a cost. If marsupials followed a shared corridor to Australia, then that corridor should show up somewhere: in Asian fossils, in surviving Asian relicts, in anything. Section 5 counted what is there. The defence that saves the probability calculation is the same defence that predicts a fossil trail, and the trail is not there.

6.3a Oard answered him, and the exchange is the most useful thing in this literature

Arment's paper did not sit unopposed. Oard replied in 2022 with a paper whose title asks the question directly, "Australian marsupials: there and back again?", and with a companion on when and how they migrated.²¹⁷

His line of defence is not to dispute the counts. It is to dispute what stratum they sit in. If the Flood/post-Flood boundary is where he puts it, and if particular Australian sites are Flood deposits rather than post-Flood ones, then the genera Arment counts as crossing a boundary never crossed anything. He writes that one site "is likely from the Flood and simply part of the deposition of marsupials and other mammals" during it.⁷

That is a legitimate move and it is also the move Arment predicted. His paper says the late-boundary proponent is left with two options, and neither is comfortable: "either each genus is its own biblical kind (contrary to what most creation biologists would accept), or they have to discard parts of the stratigraphic record as incorrectly identified in order to fit the data to their model."³ Oard has taken the second.

Nobody outside the movement is going to adjudicate this, and nobody needs to. What matters for an outside reader is that the two best-informed young-earth workers on the subject disagree about where the Flood ended, that the disagreement is empirical rather than doctrinal, and that each has published the data the other has to answer. That is how a research programme is supposed to look, and it is rare enough in this literature to be worth saying.

6.4 The endemism control

Arment's argument depends on marsupials having unusually high continental endemism. Measured here, 261 of 278 crown-marsupial genera in PBDB occur on exactly one continent, or 93.9%. (Counting Antarctica and New Caledonia as separate regions rather than dropping them gives 260 of 280, or 92.9%. The article uses the six-continent version throughout, so that it matches the control groups below.)

That number needs calibration, and the calibration cuts both ways.

Measured here.¹⁶

High single-continent rates are partly an artifact. A genus known from one quarry is single-continent by construction, and most fossil genera are known from very few sites. The skeptic who says "of course they are endemic, that is what a sparse fossil record looks like" has a point.

The point is not enough. Carnivora, at 68%, and Proboscidea, at 60%, are sampled from the same record with the same sparsity. What separates them from marsupials is that they actually crossed land bridges, repeatedly, and left the fossils to prove it. Mammoths and their relatives are the textbook Beringia crossers, and they have the lowest endemism of any group tested.

Of the multi-continent marsupial genera, most straddle pairs that were joined by land: Australia and New Guinea (Dorcopsis, Kolopsis, Nototherium, Protemnodon, Thylacinus, Zygomaturus), or Central and South America (Caluromys, Chironectes, Metachirus, Philander). Three more link South America to Antarctica (Derorhynchus, Pauladelphys, Woodburnodon), which is the Antarctic route of section 5.5. Exactly two span an ocean between distant continents, Peradectes and Peratherium, and both are northern-hemisphere stem groups, not marsupials in the sense the argument needs.

7. Two quotations that need auditing

Both are used to suggest that biogeography does not really support evolution. Both are accurately transcribed and correctly located. Neither, read whole, says that.

7.1 Nelson and Platnick

Statham writes: "Perhaps the truth is closer to the view expressed by Drs Gareth Nelson and Norman Platnick of the American Museum of Natural History, who maintain, 'biogeography (or geographical distribution of organisms) has not been shown to be evidence for or against evolution in any sense.'"¹

The quotation is verbatim and it is on page 223 of Systematics and Biogeography: Cladistics and Vicariance, as cited. Both were checked here against the book, and the leaf-to-folio map confirms the page.²²

Here is the paragraph:

Evolution thus seems to depend upon the notion that cladogram and tree are in fact bound as one, as revealed by actual evidence in hand. Our considerations suggest that relevant evidence consists only of agreement of cladograms in the geographical dimension, agreement that is by no means abundantly available at present. In fact, this evidence is notable by its very scarcity. Historically speaking, this evidence seems to have been taken for granted, which is understandable enough if only because cladograms and trees were only recently distinguished as different concepts. We conclude, therefore, that biogeography (or geographical distribution of organisms) has not been shown to be evidence for or against evolution in any sense. The significance of biogeography has been merely that biogeography has raised the possibility of agreement between cladograms as considered in the geographic dimension, a possibility that has been little studied, but one worthy of further investigation.²²

Nelson and Platnick are cladists arguing a technical point inside evolutionary biology. Their claim is that the only evidence which can bind a cladogram to a genealogical tree is geographic congruence between cladograms, that such congruence had rarely been tested in 1981, and that it deserves study. The sentence that follows the quoted one says the subject is "worthy of further investigation." Statham stops one sentence early.

There is a second problem. Congruence between cladograms in the geographic dimension has been studied a great deal since 1981, which is what the last forty years of molecular phylogeography consists of. Citing a 1981 complaint that a test had not yet been run, without saying that it has since been run repeatedly, changes what the reader takes away.

7.2 Moody

Statham quotes Paul Moody's Introduction to Evolution on natural rafts, citing the 1953 Harper edition, page 262.¹ The passage is real. It was located here in the third edition of 1970, on printed page 298:²³

In times of flood large masses of earth and entwining vegetation, including trees, may be torn loose from the banks of rivers and swept out to sea. Sometimes such masses are encountered floating in the ocean out of sight of land, still lush and green, with palms 20-30 ft tall. It is entirely probable that land animals may be transported long distances in this manner.²³

The wording differs slightly from Statham's, as it should across seventeen years and three editions, and the metric conversions in his version are his own additions. Nothing wrong there.

The problem is the sentence immediately before it, which Statham does not quote:

land mammals are few on oceanic islands, such reptiles as lizards and geckos are common. For larger land animals, therefore, some means of transportation across the ocean itself must have been operative.²³

Moody is not describing a general dispersal engine. He is in a chapter titled "Evolution as Seen in the Geographic Distribution of Animals: Oceanic Islands," and he is explaining why oceanic islands have lizards and geckos and hardly any land mammals. Rafting is his explanation for the depleted fauna, not the full one. The quoted paragraph is being used to support the opposite of the argument it appears in.

Moody also, four pages earlier, states the criterion this article measured in section 3, and states it in the same terms: continental islands sit on shelves where "the ocean between a continental island and the adjacent continent is shallow, 100 fathoms (600 ft) or less in depth," and "if the land were raised or the sea level were lowered by 100 fathoms, the British Isles would be connected to Europe and Japan would be connected to Asia."²³ Measured here: Britain to Europe, 30 m. Japan to Korea, 126 m. Both under his 183 m threshold, which is a pleasing check on a 1970 textbook and on the method.

8. Where the model is strong

A verdict that only lists failures is not an audit.

Speciation from a small founding stock is the model's best component, and mainstream biology agrees with it. Nobody now claims Noah needed 250 kangaroo species. Rapid radiation from few founders is documented everywhere, and Hawaiian Drosophila and Galápagos finches are the standard examples that creationists and evolutionists both cite. Statham is right that biogeography provides "strong evidence for the process of speciation."¹

Rafting is real, well documented, and the creationists cited it before it was fashionable again. Wise and Croxton's paper is 2003. De Queiroz's dispersal-revival review is 2005. They were not behind the field.

Their headline empirical claim checks out. Wise and Croxton report that Thorne found "78.4% of seed plant families have markedly disjunct distributions whereas this is true of only 24% of the genera and less than 1% of the species."² Thorne's own abstract, read here: "Among the Angiospermae (the flowering plants), approximately 78 per cent of the 324 families, 24 per cent of the genera, and 1 per cent of the species are widely disjunct."²⁴ Two small slips. Thorne says angiosperms, not seed plants, and gives 78 per cent where they give 78.4. Neither changes anything. The gradient is real, they reported it accurately, and it is a genuine pattern their model was built to explain.

The eighteen tests are the right way to do this. Publishing a model with a list of ways to kill it is better practice than most of what surrounds it. Several of the eighteen are genuinely checkable: current-direction versus phylogenetic polarity, endemism versus current-landing sites, disjunction frequency versus raftability. None of them has been seriously attempted in the twenty-three years since.

The Antarctic route is a better idea than the Asian land bridge, and it came from within the movement. Arment reached it by following his own data instead of the literature's habits, and he was right that "the creationist literature skews towards marsupial migration to Australia via an Asian land bridge."³ The census in section 5 says that skew is indefensible and his correction is sound. It still leaves him with a few centuries in which to do all of it.

Late Flood-boundary models are in worse shape than early ones, and a creationist demonstrated it. That is the single most useful result in this literature, and it did not come from a critic.

9. Objections and responses

"The fossil record is too incomplete to argue from absence." Generally true, and it does not apply here. The argument is not that Asia has no Pleistocene fossils. Asia has an enormous Pleistocene mammal record. The argument is that within that record, sampled by the same excavations and preserved in the same sediments, marsupials are absent while everything else is present. Absence inside a well-sampled interval is evidence. Absence from an unsampled interval is not.

"Bathymetry today does not tell you about bathymetry then." Over hundreds of millions of years, correct, and section 3.5 concedes it. Over 4,500 years it is not correct, and the young-earth model is the one that needs the change. A model claiming rapid post-Flood subsidence of the Banda Sea sills owes a mechanism, a rate, and an explanation for why the drowned surface left no terrestrial sediment.

"God guided the animals." This is available and it is not a scientific claim, so it cannot be tested and it should not be offered as evidence. Arment says the same thing about a related move, calling appeals to an innate homing beacon or divine guidance "untestable and, in the latter case, simply God-of-the-gaps theorizing."³ Any creationist who reaches for it should notice that they are conceding the empirical argument.

"Post-Flood catastrophism destroyed the intermediate fossils." Then it destroyed them selectively. It removed the marsupials from a two-thousand-kilometre corridor and left the deer, hyenas, elephants and hominins in the same beds.

"Evolutionary biogeography needs sweepstakes dispersal too, so both models are equally ad hoc." The premise is right and the word "equally" is wrong, in the creationist model's favour. This deserves a number rather than a debating point.

Mainstream biogeography needs four to six separate transatlantic crossings by rodents, monkeys, tortoises, worm-lizards and skinks,²⁵ over something like 1,450 km of open ocean.¹¹ De Queiroz's summary of why to believe it is candid to the point of being disarming: "given enough time, many things that seem unlikely can happen."⁹ Discovery Institute writers press exactly this, and Günter Bechly's version is the best-aimed: "It looks like there was some very busy ocean travel going on in those times, which suddenly stopped as soon as humans could have observed and recorded it."²⁶

How far does the creationist model actually need an animal to raft? Measured here, by the same method and on the same bathymetry as the sill table, but asking a different question: on the easiest island-hopping route from the Sunda shelf to the Sahul shelf, how long is the single longest stretch of open water?

Wallacea legs measured here;¹⁰ Atlantic figures as sourced below.¹¹

A note on that Atlantic number, because this article got it wrong once. Vidal and colleagues write that the two continents were then "between 3500 and 6000 km" apart.²⁵ That cannot be the crossing. The Atlantic has only widened since the Cretaceous, the narrowest modern gap between Brazil and West Africa is about 2,900 km, and at 35 to 40 million years ago it was narrower still. Their figure looks like a general ocean width rather than the shortest route. The figure the primary literature actually uses for the monkey crossing is around 1,450 km, at the Eocene-Oligocene sea-level lowstand.

So the two models are not asking for the same thing, and the gap between what they ask is a factor of roughly seventeen. That is a bit over one order of magnitude, not two, and it is still decisive for this objection. A critic can consistently accept an 85 km raft trip and reject a 1,450 km one. Raft survival is not linear in distance, and that is not this article's assertion. Alfred Rosenberger and Mark Hubbe, writing in Science in reply to the Peruvian monkey paper, put it in one sentence: "the odds of survival across voids of oceanic scale do not decrease linearly with distance, but rather exponentially."²⁷ Eighty-five kilometres on a two-knot current is about a day at sea. Fourteen hundred and fifty is about two and a half weeks.

Worse for the tu quoque, the creationist crossing is shorter than one that has actually been watched. The Anguilla iguanas covered 320 km, which is close to four times the Wallacea gap. Rafting across Wallacea is not an extrapolation beyond the observed record. It sits comfortably inside it.

This article got that wrong in an earlier draft, which called the two sets of inferences "identical" and accused both wings of inconsistency. They are not identical, the accusation was unfair, and the measurement above is what settles it. On this specific sub-question the creationists have the better of the argument, and Arment is entitled to cite the transatlantic crossings as supporting his own much shorter one³ while doubting the long ones.

Two things survive from the original objection, and they are the ones that matter.

Time, not distance, is the binding constraint on the young-earth side. An old-earth model can wait tens of millions of years for one lucky 4,000 km crossing. A 4,500-year model does not need a longer crossing, it needs many successful short ones, all in the same direction, within a few centuries, carrying the founding stock of a continental fauna, and then it needs them to stop.

And Bechly's "suddenly stopped" is not true, though the true version of it is sharper. Iguanas were watched arriving on Anguilla in 1995, and 289 living Japanese species were recovered off tsunami debris in North America between 2012 and 2017.¹⁸ Rafting did not stop. The reviewed reports do not document a living mammal crossing the proposed open-ocean routes. That is a real gap in the mainstream transatlantic case, and it is also a gap in the creationist Wallacea case, because marsupials are mammals too.

This is treated at length in the companion article on evolutionary biogeography, including where the mainstream account is genuinely weak.

"You have tested Statham and Wise, not the whole field." True. There is no other full-length creationist model of global biogeography in print. If one exists that this article missed, it should be tested next.

10. Verdict

The post-Flood dispersal model is a real model with real mechanisms, and two of its four mechanisms hold up. Rapid speciation from few founders is well supported. Oceanic rafting is documented, quantified, accepted in the mainstream literature, and asked to cover a distance, measured here at 78 to 110 km, that is shorter than a crossing biologists have actually watched happen.

The other two fail at the point of use. Land bridges do not reach Australia, Madagascar, New Zealand, Hawaii or the Greater Antilles, and the measurement is not marginal: 1,264 m against a maximum observed drawdown of 134 m, and against the creationist model's own 50 m. Oard reached that conclusion before this article did. Human transport after Babel cannot account for a fauna that predates every human artifact in the same beds.

So the model's weak point is not where its critics usually put it. It is not the ocean. Eighty-five kilometres is a short trip and the mechanism to make it is real. The weak points are two others, and they are both about the record rather than the mechanism.

The trail is missing. Not thin. Missing. Zero marsupial fossils in Asia after the Miocene, against 1,219 in Pleistocene Australia alone, in the same beds and the same excavations that yield everything else.

And the pattern is one-way. Rafting across an 85 km gap should be leaky in both directions and repeatedly. What Wallace's Line actually shows is a sharp, near-total faunal break maintained across that same short gap for as long as the record runs, which is the opposite of what an efficient mat-rafting corridor predicts. Oard sees this and uses it against the land bridge: "if there was such a land bridge, the fauna and flora would be more homogenous between Sunda and Sahul."⁶ The same sentence, unchanged, is an argument against efficient rafting.

The right conclusion is the one the movement's own best worker on this problem reached from the other direction. Arment's paper closes a door on late Flood boundaries and opens a narrower one, through Antarctica, for early ones. That narrow door is where the remaining creationist work on biogeography should go, and it should go there with data rather than assurances. Wise and Croxton asked for exactly that in 2003 and nobody answered.

Stating an unsolved problem plainly costs less than defending a solution that a first-year biogeography student can take apart with a depth chart.

Sources