The Creation Model of Rapid Speciation

"All the required genetic variants are present from the first generation, and are already present at high frequency, enabling very rapid selective progress." — Sanford, Carter and colleagues, International Conference on Creationism, 2018¹

"What mechanism caused such a proliferation of form and then shut it down is perhaps the most exciting unsolved mystery of creationist biology." — Kurt Wise, 1994²

Summary

Young-earth creationists hold that a few thousand animals left the Ark about 4,400 years ago and that today's land vertebrates descend from them.  The published lists put 628 living kinds aboard, against 39,586 living species.  So the model needs new species, and needs them fast.  This article states how each creationist author says it happened, in that author's own words, and then tests each account against the evidence gathered in the sibling articles.  Changes are scored as gain, loss, modification or transfer, as defined in Measuring Biological Information.

1. There is no single creation model. There are at least four, and they contradict each other. Robert Carter, Nathaniel Jeanson and John Sanford say the first animals were created carrying variety, and species formed as it was sorted. Todd Wood says that mechanism is too slow and proposes designed mobile DNA. Randy Guliuzza says organisms adjust themselves and rejects natural selection. Kurt Wise called the mechanism an "unsolved mystery" in 1994, and a writer for the organisation Wood founded said in 2025 that it is still to be found.
2. The central mechanism is well supported. Sticklebacks change armour within a decade, cichlid species were sorted from a hybrid mixture, and about 350 dog breeds arose in about 150 years. On the mainstream clock, Lake Nabugabo's cichlids became distinct-looking species in about 4,000 years. In almost every case with known genetics the change ran on variation already present; the peppered moth is the exception, a new mutation. The mainstream reviews say so in plain words. Here the creation model predicted something true, and said it before the genomes were read.
3. The required rate is fast but not absurd, and an earlier figure on this site overstated it. Counted per line of descent, the average kind must double its species about every 740 to 790 years. The fastest radiation known, Lake Victoria's cichlids, doubles every 1,600 to 1,800 years on its radiocarbon clock. That is a gap of about two. The Ark article's "300 times" compared species per year across unlike numbers of lineages. Jeanson's own linear model, which is not that error, puts the hardest demand at the start: a first split within decades, or within a few years for the mouse family.
4. The rate is absurd in one published case. Todd Wood's own example needs 150 horse species in about 400 years. That is one doubling every 55 years, about 30 times faster than Lake Victoria.
5. The variation is dated far older than 4,400 years. On measured mutation rates the reused stickleback allele is about 1.9 million generations old. On a calibrated mitochondrial clock the cichlid parent lineages split 1.6 to 5.8 million years ago. The creationist answer is that the differences were created. That answer makes a testable prediction about the kinds of differences found. Nobody has tested it. Between species the gap is wider. Mouse and rat, one kind on the creationist list, differ at about 17 percent of positions. Mutation since the Flood would need about 3,600 times the measured mouse rate, so Jeanson assigns the difference to created variety, which requires a founding pair about twice as variable as the most variable animal ever measured.
6. Two animals carry at most four versions of a gene. Wood and Jean Lightner both accept this ceiling, and both concede that living kinds carry far more. Dogs and wolves have 90 named versions of one immune gene. The replies offered are mutation, gene conversion and designed mechanisms. The measured conversion rate is high enough to make the count. It cannot make new sequence, which gives a second test nobody has run.
7. Continuous environmental tracking failed where it was checked. Guliuzza predicted that variation assumed to be random would prove to be directed. At the stickleback armour gene, offspring were born in the plain Mendelian ratio of 1:2:1, and the frequency then moved because some fish died.
8. Sorting and loss cannot build what differs between kinds. The model accepts this. It is a model of how one kind becomes many species. It says nothing about where a kind's genes came from.
9. Faster mutation is no way out, and the model's authors do not claim it for the cell nucleus. Making the dated differences by mutation since the Flood would need rates 92 to 431 times the measured ones. At that rate a human child would carry 200 to 540 new harmful mutations. Sanford's own team says a hundred mutations a generation drives a population extinct if the mutations do harm.

Verdict.  The mechanism stands and the chronology does not.  Fast speciation from stored variation is observed, and both sides now agree on it.  The claims that the variation was created and passed through one pair about 4,400 years ago are untested where they could be tested and contradicted where clocks have been measured.

Part 1: What the model has to explain

The need, in the model's own words

The need is stated most plainly by the people who hold the model.  Kurt Wise, a young-earth palaeontologist, wrote in 1994 that land-mammal kinds commonly hold scores of species, that only a pair of each went aboard, and that ancient art shows modern forms within a few centuries of the Flood.  His conclusion:  "This means that there would of necessity have been an explosive amount of biological change in the few centuries after the Flood."²  He added that the change was "too rapid a change to have been produced by any traditional microevolutionary mechanisms (for example, mutations acted on by natural selection)".²

Todd Wood made the requirement concrete in 2003.  Creationist studies had placed every fossil horse in one kind.  Genesis has Pharaoh giving Abram donkeys about four centuries after the Flood.  "This would require the origin of 150 species of horse prior to the gift of donkeys (Equus asinus) from Pharaoh to Abram (Gen. 12:16), which took place ~400 years after the Flood."³  The year before he had written that "No evolutionist would ever believe that all of horse, cat, or camel evolution occurred in less than five centuries, but this is the true essence of diversification."⁴

Jake Ramgren, writing in 2025 for the organisation Wood founded, put it in one line:  "Creationism requires the origin of species just as much as Darwin did".⁵

How many kinds, how many species

The sibling article on the Ark's logistics counted both sides of the ledger from the source databases.  The kinds papers published in Answers Research Journal sum to 628 living kinds.  Jean Lightner's mammal paper gives 137 of them.⁶  Against these stand 39,586 living species of mammals, birds, reptiles and amphibians, 6,791 of them mammals.⁷  The average kind has therefore become 63 living species.  Fossil species from rocks laid down after the Flood add to every total, and the horses show by how much.

The species are not spread evenly.  Carter notes that of the mammal families "some have upward of 400 species, but most have less than 20."⁸  The current database gives the New World rats and mice, the family Cricetidae, 882 living species, and the Old World rats and mice, Muridae, 865.⁷  Lightner lists each family as one kind, while allowing that "the level of the kind could easily be at this higher level", the superfamily.⁶

The rate per lineage

Species divided by years is the wrong measure, because a group with many lines of descent makes more species a year than a group with few, at the same pace in each line.  The standard measure is the doubling time.  Ole Seehausen used it for cichlids in 2002:  time multiplied by the natural logarithm of 2, divided by the natural logarithm of the ratio of final to starting species.⁹  The table applies it to the model, with 4,400 years since the Flood.¹⁰

Now the comparison.  The fastest radiations in the literature are all inferred from dated lakes and family trees.  None was watched.  The sibling article on cichlid radiations gives Lake Victoria a doubling time of 1,628 to 1,833 years on the radiocarbon date for the lake's refilling.  It gives the pupfish 1,893 to 3,155 years, the ciscoes about 6,000, and the cichlids of crater Lake Apoyo at most 1,818.^{9 10}

So the average kind must diversify about twice as fast as the fastest radiation known, and the rodent families about four times as fast.  That is a real demand.  It is a demand on every kind at once, where Lake Victoria is the single fastest case known.  But it is a factor of two to four, and the model has its own reply:  on a 4,400-year clock Lake Victoria itself needs only 491 to 552 years, which is the rodent rate.  The model is at least consistent with its own best example.

Two limits on the table.  The figures are net rates, as Seehausen's are, so every line that died out had to be replaced by another split and the real demand is higher.  And in rates per line per century the table runs from about 0.09 for the average kind to 0.15 for the rodent families, against about 0.04 for Lake Victoria.¹¹  Birds give the same order.  The Darwin's finches article finds that Jean Lightner's sparrow-finch kind of more than 1,000 species needs each line to split about once every 440 years, the rodent figure.

Wood's horses are different.  One doubling every 55 years is about 30 times faster than Lake Victoria on the radiocarbon clock, and nine to ten times faster than Lake Victoria on the model's own clock.  Nothing inferred or observed comes near it.  Wood drew the same conclusion:  "Since we see no ‘speciation’ like this today or in historical times, we can infer that the mechanism of intrabaraminic diversification is qualitatively different from modern speciation."³

One thing can be said for the horses, and it does not help much.  Fossil species are named from bones, and ancient DNA has merged some of them.  Peter Heintzman and colleagues found that the stilt-legged horses of Ice Age North America, named as several species, were "only a single species".¹²  If the true count were a fifth of Wood's 150, the doubling time would be 82 years instead of 55, still about 20 times faster than Lake Victoria.¹³  Because the count enters as a logarithm, the 400-year window drives the result, not the number of species.  The horse series article finds the family a branching bush of about 51 genera, so the splits are real and not just one line renamed at each step.

One correction to a figure in circulation follows from this.  The Ark article on this site said, until it was corrected, that the model needs about 1.5 new species a year, "about 300 times the fastest measured vertebrate radiation".  That divides species by years for groups with very different numbers of lineages.  Per lineage the gap is about two.

Jeanson's figures look like the same mistake and are not.  His 2015 paper says that "one speciation event every 200 years would produce about 21 species by today."¹⁴  That is a constant number of new species per kind per year, and it is his hypothesis, not an oversight.  He fitted it to mitochondrial family trees and drew the consequence for each line himself:  "the per species speciation rates were (and are) declining", because a fixed number of new species a year is shared among more and more lines.¹⁴  Carter's picture is the doubling one:  "the kind can divide into two species, which divides into four, etc."⁸  The two pictures need the same total and spread it differently over time.

The linear picture puts its hardest demand at the start.  Jeanson gives the cattle family, Bovidae, "one new species every 31 years".¹⁴  Held from the Flood, that rate splits the founding line within about 31 years, splits both daughters within the next 62, and passes eight species in about 220 years.  For Muridae, with 716 species in his table, the first split comes about six years after the Flood.¹¹  A first doubling in 31 years is 50 to 60 times faster than Lake Victoria's doubling on the radiocarbon clock, and 16 to 18 times faster on the model's own clock.  One in six years is 270 to 300 times faster, or 80 to 90 times.  The demand falls when the founding population was smallest.  The doubling picture asks less at the start and more at the end.  No rate inferred from a dated lake matches either.

In generations the demand looks different again.  Carter observes that the species-rich families "tend to include the smaller species, with high reproduction rates, short generation times, and large populations".⁸  A mouse kind that doubles every 450 years has hundreds of generations for each split.  The sibling article on observed speciation records a finch lineage isolated by song and size within three generations.¹⁵  For small animals the generations are there.  For horses they are not.

Part 2: The models in their own words

Four accounts are in print.  Their authors use similar words for different things, so each is quoted with its own term and its own definition.

Created heterozygosity

Heterozygosity means carrying two different versions of a stretch of DNA, one from each parent.  The claim is that the first animals of each kind were made that way at very many places in the genome.

Jeanson and the astrophysicist Jason Lisle gave the technical statement in 2016 and named it the Created Heterozygosity and Natural Processes hypothesis.  It "proposes that diploid individuals were created heterozygous, and that natural processes since this event (including recombination, gene conversion, mutation, natural selection, etc.) have distributed and/or added to the original created genetic diversity, thus producing the genotypic and, consequently, phenotypic diversity we observe today."¹⁶  Jeanson's book Replacing Darwin puts it in a sentence:  "I claimed that the ancestors to modern species were created with heterozygous chromosome pairs."¹⁷

Carter of Creation Ministries International states the same idea as the common view:  "The prevailing idea within creationist circles is that God created distinct “kinds” that were front-loaded with a large amount of genetic diversity that could then be used as the kinds turned into our modern species."¹⁸  His recipe for a new species has three parts:  "All we need is an initially high genetic diversity, a way to scramble that diversity, and the isolation of subpopulations."⁸

The reason given for speed is the one in the first pull quote above.  Sanford, Carter and four colleagues wrote that "no extended “waiting time” is required", because the variants are present from the first generation and already common.¹

Three features of this model matter for what follows.  First, its authors say that mutation and selection alone cannot do the work.  Jeanson and Lisle computed that selecting a million variants one at a time would need a cycle of population growth and crash "once every ~1.6 days", and called it not plausible.¹⁶  The engine they propose is sorting by isolation and chance.  Second, the direction of change is loss.  Carter:  "Natural selection doesn’t create genes, but instead culls them."⁸  Third, the idea is old.  Wood's 2002 survey traces it through six creationist authors from 1974 to 1997.⁴  It was in print decades before any genome was sequenced.

Designed mechanisms of change

Wood examined created heterozygosity in 2002, under its older name of heterozygous fractionation, and found it wanting on two counts.  Breeding makes varieties and not species:  "the changes that result from heterozygous fractionation would probably not yield true species in the time available."⁴  And two animals cannot carry enough versions of a gene, a point taken up in Part 4.

His own proposal is that mobile stretches of DNA were designed to move to particular places and switch genes on or off.  He named them Altruistic Genetic Elements.  They "fulfill three criteria for explaining post-Flood diversification: 1) they permanently alter the genome, 2) their alterations can be gene-specific and beneficial, and 3) their beneficial activity was concentrated in the past."⁴  He saw what the third clause costs.  Elements "inserting at random positions in the genome would induce changes at the same frequency as other mutations, yielding little benefit to the host organism."³  The model needs insertions aimed at the right places, and needs them to have stopped.

A second version appeared in the Journal of Creation in 2009 under the name Peer Terborg.  Created kinds "were frontloaded with baranomes: that is, pluripotent genomes with an ability to induce variation from within."¹⁹  The agents are the same mobile and repeated sequences, renamed variation-inducing genetic elements.  The hypothesis "may be a framework to understand the origin of diseases and explain rapid speciation events through facilitated chromosome swapping."²⁰  The author adds a caution:  "Whether the ultimate effect on the phenotype of the organism can be predicted, however, remains to be established."²⁰

Jean Lightner, a veterinarian who wrote the mammal kinds paper, argues from patterns.  Her term is "directed" mutation:  "many genes are designed to allow for potentially useful mutations and some mutations appear to be directed."²¹  From the dog family she concluded that "certain genetic components appear designed to change and numerous designed mechanisms may be involved in driving many of these changes."²²  In 2013 she named ordinary recombination as one such mechanism, and held that "the genome must contain biological information that allows it to induce variation from within."²³

None of these papers reports a case in which a designed element was watched making a useful change on cue.  None gives a rate.

Continuous environmental tracking

Randy Guliuzza of the Institute for Creation Research proposes something different in kind.  His 2018 paper with Phil Gaskill reviewed 342 journal articles and offered a hypothesis:  "organisms actively and continuously track environmental variables and respond by self-adjusting to changing environments".²⁴  Organisms are said to carry sensors, logic and actuators, as an engineered tracking device does.  The paper reports that "Adaptation happened largely through regulated gene expression and not gene inheritance, per se."²⁴

Guliuzza rejects selection by the environment outright.  "But since a real “selection event” is lacking, the only place selection is happening is in their own minds."²⁵  His evidence from the institute's own laboratory is blind cavefish regaining pigment:  "within a month, cavefish can regain pigmentation when placed in simulated river conditions, and surface fish lose pigment when placed in simulated cave water.  It’s significant to note that no fish died."²⁶

The model makes one explicit prediction about inheritance:  "we predict further investigation will demonstrate that many variations which are assumed to be—but never documented to be—random in nature were never random at all but directed."²⁴  With Gaskill he applied the model to sticklebacks, whose traits he described as "produced “on the fly” by changes to the “instruction code”".²⁷  Part 4 tests that.

Mutation that is not random

In January 2022 Grey Monroe and fifteen colleagues reported in Nature that new mutations in the plant Arabidopsis thaliana are rarer where they would do most harm:  "mutation frequency is reduced by half inside gene bodies and by two-thirds in essential genes."²⁸  They tied the pattern to chemical marks on the proteins that package DNA, which draw repair enzymes to important genes.  Their conclusion was that this bias "reduces the occurrence of deleterious mutations in Arabidopsis, challenging the prevailing paradigm that mutation is a directionless force in evolution."²⁸

That is the whole claim.  Mutations are fewer in important genes.  The paper does not say that a plant makes the mutation it needs, and it does not mention speciation.

Two creationist readings followed.  Jeffrey Tomkins of the Institute for Creation Research called the pattern "an ingenious scheme to benefit the genetic integrity of the plant" and drew a model from it:  living things "had a certain amount of genetic variation built into them at creation", and also "had mechanisms built into them to facilitate novel genetic variation as part of their adaptive design."^{29 29}  Bruce Lawrence of Creation Ministries International read it more carefully.  The marks do not judge a mutation.  "They merely detect a copying error and send out chemical signals to bring in repair machinery."³⁰  His conclusion was the reverse of Tomkins's second clause:  "The genome appears to be designed to resist the major evolutionary changes that are needed to create the extreme diversity seen among living things."³⁰

Lawrence has the paper right.  A mechanism that suppresses change in genes cannot be the source of new versions of genes.  Monroe's result, if it stands, gives the rapid speciation model nothing.

It may not stand.  Haoxuan Liu and Jianzhi Zhang looked for the trend in yeast and in humans.  "The yeast data do not exhibit this trend, whereas the human data show an opposite trend that disappears upon the control of potential confounders."³¹  They then checked Monroe's mutation counts against the studies that produced the raw data.  In one set "The original authors identified 17 mutations, but Monroe et al. reported 773,141 mutations by reanalyzing the published sequencing reads, a 45,479-fold difference".³¹  The extras, they concluded, "are likely sequencing errors."³¹  Long Wang and colleagues, in Nature itself, found "abundant sequencing and analysis artefacts" and judged that Monroe's "claims are not robustly substantiated."³²  Monroe's group replied that they "agree that there are issues and inherent uncertainties with somatic mutation calling", and stood by the pattern.³³  Tomkins's "million-plus mutations" are those disputed calls.  The dispute is open.  Neither side of it claims that organisms write their own useful mutations.

The models contradict each other

Jeanson tested the early-burst view against mitochondrial DNA in 2015 and "found no evidence in support of the early episode hypothesis".¹⁴  Wood had already judged Jeanson's mechanism inadequate in 2002.  Guliuzza rejects the selection that Carter and Sanford rely on.  Ramgren, surveying the field in 2025, wrote that the question "still requires young-earth scientists to discover what mechanism can cause drastic macroevolutionary change in only a few thousand years."⁵  A critic who answers "the creation model" has answered one author.

Part 3: What the evidence supports

The sibling articles examined the best cases of fast change in animals and plants.  In almost every case where the genetics is known, the change drew on variation that was already there;  the peppered moth ran on a new mutation.  The table gives each case, what was watched, and how the site's taxonomy scores it.

Each sibling gives the limits.  In sticklebacks nothing new was watched arising.  No cichlid species was watched forming.  Many dog variants arose by mutation in dogs long before the breeds.  New species have been watched forming only in plants, by a doubling that adds no gene.

Lake Nabugabo is the case closest to the model's own time scale, and it was dated by the mainstream.  Humphry Greenwood of the British Museum described its fishes in 1965.  "The age of Lake Nabugabo has been estimated at approximately 4,000 years," from radiocarbon on charcoal in an old Victoria shoreline at the height of the sand bar that closed the lagoon.⁴¹  Five of its six Haplochromis species live nowhere else, and "the morphological gaps between any one Nabugabo species and its counterpart in Victoria are as great as those between any two related Haplochromis in Lake Victoria."⁴¹  He called it "a text-book case of geographical isolation resulting in speciation."⁴¹  So on the mainstream clock a cichlid line can become a distinct species in under 4,000 years.  The limits are Greenwood's own.  He wrote:  "In relation to the species of Lake Victoria, those of Nabugabo must be considered inferential".  No one had tested whether they would interbreed with their Victoria counterparts.⁴¹  Each line also budded once, so the lake shows at least one new species per line in 4,000 years.  The model needs a split every 450 to 790 years, five to nine times that, and Nabugabo cannot say how much faster than 4,000 years its species formed.  The date rests on one indirect radiocarbon reading.

The Pod Mrčaru wall lizards are the reverse lesson.  Ten lizards were moved to a tiny Croatian islet in 1971, and 36 years later their descendants ate plants, had larger heads and carried valves in the gut.⁴²  The discoverers wrote that plasticity and maternal effects still had to be tested.⁴²  The same group later found the valves "not fixed features".⁴³  The Italian wall lizards article scores the case as not yet scoreable.  It is the best evidence in this series for Guliuzza's point that much fast change in form is an organism adjusting to its surroundings, which Part 4 takes up.

The mainstream literature says the same

The point is not a creationist reading laid over reluctant data.  Rowan Barrett and Dolph Schluter reviewed the subject in 2008 and led with it:  "adaptation is likely to be faster from standing variation than from new mutation, not only because beneficial alleles are immediately available, but also because they usually start at higher frequencies".⁴⁴  Standing variation is the variety already present in a population.

David Marques, Joana Meier and Ole Seehausen went further in 2019.  They opened with the difficulty:  "Speciation is often thought of as a slow process due to waiting times for mutations that cause incompatibilities, permit ecological differentiation or assortative mating."⁴⁵  Their answer was "re-assembly of old genetic variation into new combinations facilitating rapid speciation and adaptive radiation."⁴⁵  Such a mechanism works fast, they wrote, "as there is no waiting time for relevant mutations".⁴⁵

Set that beside Sanford's team a year earlier:  "no extended “waiting time” is required."¹  The two camps describe the same mechanism in nearly the same words.  The creationist version is the older.  Wise wrote in 1994 that in the breeding of dogs and pigeons the variety "was always extant (sometimes largely in latent form) and merely expressed".²  Arthur Jones, a creationist who wrote his doctorate on cichlids, said in 2001 that their diversity "was actually produced by the endless permutation of a relatively small number of character states".⁴⁶  Meier's team found in 2017 that cichlid speciation "has commonly been associated with sorting of alleles brought together in the radiation ancestor".³⁶

This is a real success for the model, and the site says so.  It comes with a cost that Part 4 takes up.  The same two reviews that confirm the mechanism also date the variation.  Barrett and Schluter report that the stickleback allele "originated >2 million years ago."⁴⁴  Marques and colleagues say the variants are older than the species, "sometimes by orders of magnitude."⁴⁵

One mechanism that does not carry over

Polyploidy makes a species in a generation, and creationist writers cite it.  It should not be offered as a model for animals after the Flood.  Jerry Coyne and Allen Orr write that its rarity in animals "probably represents “the greatest known difference between the evolutionary patterns in the two kingdoms”", quoting Theodosius Dobzhansky.⁴⁷  They also reviewed the laboratory experiments in which flies were selected for many generations.  Isolation appeared quickly, "Although in no case was reproductive isolation complete".⁴⁷  Fast sorting makes forms fast.  Whether it makes species that stay apart is a separate question, and it is the one Wood raised about dogs.

Part 4: Where the model is in trouble

Each problem is stated at full strength, followed by what creationist authors have published in reply.

The age of the variation

The alleles that fast change reuses can be dated, and the dates no longer rest only on assumed calibrations.

In sticklebacks the two versions of Eda differ at 1.938% of the positions where a change does not alter the protein.  Pamela Colosimo's team dated the split at "more than two million years ago" by comparison with other fish.⁴⁸  In 2025 the mutation rate of the threespine stickleback was measured directly, by sequencing parents and offspring:  "µ = 5.11 × 10^{−9/bp/gen".[}zhang-2025:rate]  The sticklebacks article divides one by the other and gets about 1.9 million generations.⁴⁹  Measuring the rate made stickleback dates older, which is the reverse of what happened with human mitochondrial DNA.

In cichlids the two lineages that mixed to found the Lake Victoria flock split "~1.6–5.8 million years ago".³⁶  In Lake Malawi a rate measured in cichlid families gives splits hundreds of thousands of years deep.  In Tibetans the high-altitude version of EPAS1 sits on a stretch whose tree runs back about a million years.⁴⁹

The reply.  The differences are not mutations.  They were written into the first animals.  Jeanson:  "the vast majority of these differences are not the product of error, but of deliberate design."  He suggests about 50 million created differences between the two DNA copies of the first cat.⁵⁰  On this view a mutation clock applied to created differences measures nothing, and the millions of years are an artefact.

That reply is coherent.  It is also nearly free of content as it stands, because any amount of difference can be assigned to creation.  Carter says as much:  God "could have engineered any amount of genetic diversity into the genomes of the created kinds."¹⁸  Jeanson and Lisle's stated prediction does not help.  They "anticipate that constant rates of nuclear DNA mutation will be found to be insufficient to explain extant genetic diversity" on a 6,000-year scale.¹⁶  The long-age model predicts the same observation.  A measured rate too slow to make the differences in 6,000 years is exactly what "two million years old" means.

What the reply does predict.  Mutation leaves a signature, because some changes are far commoner than others.  The stickleback pedigrees show a "typically high proportion of C > T mutations", and places where a cytosine sits beside a guanine mutate faster than the genome average.⁵¹  If the differences between the two Eda versions accumulated by mutation, they should show that signature.  If they were composed, they need not.  The study catalogued 115 new mutations, so the signature is known for that fish.⁵¹  The old differences are published.  The comparison has not been made by anyone on either side, as far as this survey found.  It is the most direct test of created variation available, and the sticklebacks and cichlids articles both arrived at it independently.

A designer could of course compose differences that imitate the mutation signature.  A model that expects that has stopped predicting anything.

How far apart the species of one kind are

The age of the variation inside one species is the smaller problem.  The larger one is the distance between the species of one kind, and Jeanson has stated the young-earth answer to it in his own words.

Lightner's list makes the house mouse and the brown rat one kind:  "Muridae (Old world mouse/rat/gerbil/whistling rat kind)".⁶  Their genomes differ at about 17 percent of positions.  Sarah Lindsay and colleagues at the Sanger Institute, who measured mutation in six mouse families, give "mouse–rat sequence divergence (1.3% and 17%)" beside the human–chimpanzee figure.⁵²

Jeanson:  "To explain why over 35 million single base pair differences separate species within the same genus (let alone species within the same family), young-earth creationists invoke the creation of heterozygosity within the first (created) members of the Murid family."¹⁷  He is right to rule out mutation.  The measured mouse rate is about 5.3 × 10⁻⁹ per position per year.⁵²  A 17 percent difference built along two lines in 4,400 years would need about 3,600 times that rate.  No creationist proposes this, and the figure is here for scale only.⁵³

So created difference has to carry the load, and the load can be measured.  Thirty-five million differences is about 1.3 percent of a mouse genome, and the family needs more than ten times that.  Sorting cannot make two lines more different than the copies they were sorted from.  So among the four copies of each chromosome on the Ark, the most different two must have differed about as much as mouse and rat now do.  Any offspring of the pair that inherited those two copies carried a 17 percent difference inside one animal.  A survey of 167 species by Ellen Leffler and colleagues found that "nucleotide diversity π ranges from 0.01% per base pair in Lynx lynx to 8.01% in Ciona savignyi", a sea squirt.⁵⁴  The mouse kind's founders would need about twice the variety of the most variable animal measured.

The reply is that created genomes need not resemble living ones, and nothing measured today rules that out.  It is the same reply as before, applied at the scale of a family:  any amount of difference can be assigned to creation.  The test also carries over.  If murid differences were sorted from four copies, then inside any stretch short enough to escape recombination, every mouse, rat and gerbil should carry one of at most four deep versions plus a light scatter of new mutations.  Nobody has looked.

Two animals, four versions

A pair of animals carries four copies of each chromosome between them, and so at most four versions of any gene.  This objection is usually put by critics.  It was put first, and best, by creationists.

Wood, 2002:  "After the Flood, the maximum number of alleles per locus for any land baramin would be four (unclean; or 14, clean).  A survey of modern allelic diversity reveals a greater diversity than just four alleles per locus per baramin for many loci".⁴  Lightner, 2009, on the dog family:  the pair on the Ark carried "up to four different versions of any particular gene", and "Today there is considerably more".²²  Her example is the set of immune genes that mark cells as self:  "As of 2006, there were 90 alleles recognized for DLA-DRB1, 22 for DLA-DQA1 and 54 for DLA-DQB1, with more expected to be discovered."²²  Her conclusion is plain:  "significant diversity has arisen since the Flood."²²

The replies.  Three are in print.

Carter's first reply is that the ceiling may not apply.  "We cannot assume God created two of every animal.  All we know is two of each were on the Ark, and we do not know how closely related the pairs were."³⁴  That is true of creation week.  It does not touch the Ark, where the text gives two of each unclean kind.  For humans he goes further:  "there is no way to say that God did not front-load Adam and Eve’s reproductive cells with many different genomes".³⁴  A pair whose reproductive cells each hold a different genome is no longer a pair in the genetic sense, and the four-version ceiling was the only thing that made the model checkable.

His second reply is reshuffling inside the gene:  "chromosomal recombination can lead to an increasing number of ‘alleles’ because the initial versions of a gene get scrambled to make more versions."³⁴  Jeanson and Lisle add plain mutation:  "within a species, mutation can generate far more than the 4–28 alleles that were brought on board the Ark."¹⁶

Lightner's reply is designed mechanisms, with gene conversion named.  Gene conversion copies a short stretch from one version of a gene onto another.  She cites a yeast study with "a mutation rate 1,000 times higher during gene conversion than the normal spontaneous mutation rate for that locus."²³

Pricing the replies.  None of these authors gives a rate for an immune gene.  Measured rates exist.  In human sperm, "about 1/10,000 sperm represents a new DPB1 sequence generated by short tracts of segmental exchange (gene conversion)".⁵⁵  A mouse study found a rate fifty times lower for one named stretch moving to one named gene.⁵⁶

Carry the human figure to the dog family, for scale only.  A population of 1,000 breeding animals makes about one new sequence every five generations.  Over the 1,467 three-year generations since the Flood that is about 290 new sequences at one gene.  A population of 10,000 makes about 2,900.  Eighty-six were needed.  Plain mutation is about 74 times rarer in a stretch that size and would supply a few dozen.⁵⁷

So on the count of versions the creationist reply holds.  The sibling article on human genetic diversity found the same for humans:  most of the thousands of named human versions were made by gene conversion and point mutation from a few older lineages.  The allele count is a weak objection and should be retired.

The strong form of the objection is different.  Conversion and recombination rearrange what exists.  They make no new stretch of sequence.  If a kind began with four versions, every version alive today should be a mosaic of those four, with a light dusting of new mutations.  Deeply different lineages should number four or fewer.  In humans the sibling article sets thirteen named lineages at one immune gene against a ceiling of four, and notes that the number of truly deep lineages is not settled.  In wolves a Croatian study of 77 animals found "13 DRB1, 7 DQA1 and 11 DQB1 highly divergent alleles".⁵⁸  It also reports allele lineages shared across species.⁵⁸  The abstract does not name the species, and if they are other members of the dog family the finding sits inside the kind.  Whether the 90 dog and wolf versions reduce to four founding sequences is a question the public data could answer.  No creationist author has asked it in print.

What a founding pair should leave behind

A founding pair keeps most of its source population's heterozygosity and loses most of its alleles.  Masatoshi Nei and colleagues showed both in 1975.  Two individuals "can be regarded to contain 75% of the total genetic variability", and yet "the average number of alleles contained in the sample is much smaller than that in the parental population".^{59 59}  So low variety today is not what the model predicts, and species with plenty of variety do not refute it.

Recent bottlenecks that lasted do show.  The northern elephant seal fell to about twenty animals in the 1890s.  Its heterozygosity dropped from "0.00142" to "0.000176" and has not recovered.⁶⁰  The wolves of Isle Royale, founded by two or three animals in the 1940s, carry "23 to 48%" of their genomes in long identical runs.⁶¹  The Ark event would not look like either.  It was one generation at a size of two, followed by fast growth, and it was long ago in generations.

What it would leave is a pattern of blocks.  Each chromosome is cut and rejoined about once per generation along a standard length.  After 1,467 generations a stretch from a founder survives unbroken for about 0.07 of a percent of that length, which in the human genome would be about 60,000 letters.⁶²  Inside any window shorter than that, all the animals of all the species of a kind should carry one of at most four founding sequences, plus rare new mutations.  The common variants should also sit near frequencies of one quarter, one half and three quarters in large populations, because that is where a pair starts them.

For humans the sibling articles Population Genetics and human genetic diversity cover the ground, and it is not repeated here.  The short version is that block counts fit a pair and the depth of difference between blocks does not.  Jeanson and Lisle tested block number for humans and found that "the predicted number of haplotype blocks overlapped the current number."¹⁶  For animals nothing has been published.  Carter explains why:  "we cannot do this for animals because we simply do not know the underlying population genetics."³⁴

There is one compartment where the model cannot appeal to created variety.  Mitochondria pass through the mother alone.  One female per kind means one mitochondrial sequence per kind.  Wood accepted this:  "any mitochondrial sequence variability must be attributed to real mutations instead of allelic diversity preserved from before the Flood".⁶³  He then measured the cat family and found differences at up to 46 of 287 positions in one gene.  His answer was "a period of rapid genetic divergence after the Flood that quickly decreased to the low mutation rates characteristic of the present."⁶³  No mechanism is proposed for the burst or for its end.

Tracking, tested at one gene

Guliuzza's claim, quoted in Part 2, is that organisms adjust themselves, that stickleback traits are produced "on the fly", and that variation thought random will prove directed.

Rowan Barrett's team ran the test in 2006 without intending to.  They put marine sticklebacks carrying one copy of each Eda version into four freshwater ponds and genotyped the next generation.  If fish altered their DNA to suit fresh water, the offspring should have leaned toward the low-armour version from the start.  They did not.  "Genotype frequencies at Eda in the F1 generation were not significantly different from the predicted 1:2:1 ratio".⁶⁴  One quarter, one half, one quarter is what blind shuffling of the parents' copies gives.  The frequency then shifted because fish with some genotypes died more often than others.  In the Alaskan lakes the low-armour form tracked the square of the allele's frequency, which is the mark of an ordinary recessive allele under selection.³⁵

Guliuzza's cavefish result does not answer this.  Pigment returning within a month in living fish, with "no fish died", is a change in the use of genes during a lifetime.  Biologists call that plasticity and have no quarrel with it.  The claim in dispute is that inherited change is directed.  At the one gene where inheritance was checked, it was not.

Tracking may hold for traits that are plastic, and the Pod Mrčaru lizards in Part 3 are the strongest candidate in this series:  a gut structure first reported as fast evolution now looks, to the team that found it, like a response that comes and goes.  It fails as an account of how armour is lost, and no creationist reply to Barrett's ratios was found.

What sorting and loss cannot build

Sorting divides what exists.  Loss removes it.  Across the sibling articles the scored changes in fast-changing animals are losses, weakened genes, extra copies and transfers.  The dog article tallies four losses, two genes turned down, four turned up and three unmeasured, with no gain.  The stickleback pelvis is lost by deletion.

The created-heterozygosity model accepts this limit and builds on it.  Carter:  "Change is (mostly) a one-way street."⁸  The model explains how a dog kind becomes wolves, foxes and jackals.  It does not try to explain where the dog kind's genes came from, which it assigns to creation.  That is a legitimate boundary.  It means the model and the mainstream account do not compete over fast speciation.  They compete over when the variation arose and how the differences between kinds were made, and the evidence in this article bears only on the first.

The limit cuts the other way too.  This series concedes that mutation sometimes makes something new:  a citrate-using E. coli, an antifreeze gene in cod, some antibiotic resistance.  The dog article found that many breed-defining variants arose by mutation in dogs and were not in wolves.  Lightner concedes it for coat genes:  "I think it is more likely that these alleles have arisen through mutation."⁶⁵  The technical models allow for this.  Jeanson and Lisle's hypothesis has natural processes that "distributed and/or added to" the created variety.¹⁶  So the created store is not the whole account even on the model's own terms.  How much of a kind's useful variety was created and how much is new has not been measured by anyone.

Part 5: Pricing the mechanisms

How much a pair can carry, and how fast it goes

A pair can carry a great deal.  Nothing in genetics limits the number of places at which two animals differ.  Jeanson suggests 5 to 10 million created differences in the first man.⁵⁰  Carter says a person today carries two versions at 4 to 5 million places.⁶⁶  The figures agree by construction, since the created number was chosen to fit.

What the pair cannot do is keep it while splitting.  Each time a new species is founded by a few animals, part of the variety stays behind.  A founding pair keeps three quarters.  The average mammal kind needs 5.6 doublings to reach today's species count, and the rodent families need 9.8.  If every new species began with a pair, the last species in the line would hold 20% of the first pair's heterozygosity after 5.6 foundings and 6% after 9.8.  With ten founders each time the figures are 75% and 61%.⁶⁷  Chance losses in small populations after each founding take more.

This is the model's internal tension.  Sorting makes species differ by fixing different versions in different lines, and every version fixed is variety spent.  The faster and tighter the founding events, the quicker the forms diverge and the sooner the store runs out.  Carter expects this:  "I fully expect that the rate of speciation has slowed down over time."⁸  The model therefore predicts that species-rich kinds hold less variety per species than species-poor kinds of similar population size.  That can be checked against published genome surveys.  It has not been.

How fast form can change

Speed of change in form is not the bottleneck.  Dogs show it.  Jocelyn Plassais and colleagues at the National Human Genome Research Institute open their 2022 paper:  "Domestic dogs (Canis lupus familiaris) are the most variable-sized mammalian species on Earth displaying a 40-fold size difference between breeds", and "the most dramatic shifts in body size are the result of selection over the last two centuries".⁶⁸  The main small-size version of IGF1, the gene with the largest effect, is the ancestral state in wild canids, so breeders sorted it and did not have to wait for it.⁶⁸  A 40-fold range of size arose inside one species in about two centuries.

The mainstream reply is that wild animals do not behave like breeders' dogs.  Josef Uyeda and colleagues pooled more than 8,000 measurements of body size in 2011, from studies of living populations, fossils and comparisons between species.  "Even though rapid, short-term evolution often occurs in intervals shorter than 1 Myr, the changes are constrained and do not accumulate over time."⁶⁹  Fast change in the wild goes back and forth.  The Darwin's finches article found the same on Daphne Major, where beak size swung with droughts and its mean in 2001 was the same as in 1973, though shape had changed.

This is the real difficulty for the model on form, and it concerns direction, not speed.  The model needs post-Flood lines to keep changing the same way for centuries, as breeders made dogs do, while measured wild populations over spans of up to a million years mostly fluctuate.  The creationist can point to the explanation Uyeda's team favours, taken from Douglas Futuyma:  "the mosaic of niches and diverse adaptive optima of wide-ranging species prevents local evolutionary changes from spreading across the entire range".⁶⁹  A kind spreading from a few founders into empty land is the case where that brake should be weakest.  That reading is reasonable and untested.  It does not turn the measured pattern into support.

What faster mutation would cost

No creationist author proposes that nuclear differences arose by fast mutation after the Flood.  The arithmetic below is for scale, and it shows why they are right not to.

The sibling articles computed the mutation rate that would make the dated differences in the time allowed.  At stickleback Eda it is 431 times the measured rate.  At Tibetan EPAS1 it is 92 to 245 times.  For Lake Malawi cichlids it is 97 to 195 times.⁴⁹

A human child carries about 2.2 new harmful mutations:  "estimates are U ≈ 2.2 for the whole diploid genome per generation".⁷⁰  The Genetic Entropy article argues that even this is hard to sustain.  At 92 to 245 times the rate, each child would carry 200 to 540.⁷¹  A stickleback would carry about 2,040 new mutations of all kinds in place of 4.7.

Sanford's team states the consequence in its own methods section.  They made the mutations in their simulations neutral because otherwise "the population will go extinct due to slightly deleterious mutations, long before the simulation is complete."¹  That was at 100 mutations per person.  A mutation rate high enough to help the chronology would end the population.

Wood's mitochondrial burst meets the same wall on a smaller genome.  The cat family's differences need about 1.8 × 10⁻⁵ changes per position per year along each line, held for 4,400 years or concentrated into less.⁷²  The Mitochondrial Eve article gives a measured human rate of about 10⁻⁶ in the fastest region of the molecule.  The gene Wood measured lies outside that region.

Objections and responses

"Created variation cannot be dated by a mutation clock, so the ages mean nothing."  Granted, if the differences were created.  The objection assumes what is in dispute.  It becomes evidence only when the differences are shown to lack the signature of mutation.  That test is available and has not been run.

"Lake Victoria proves hundreds of species can form in a few thousand years."  It shows that about 500 forms arose in under 16,000 years by radiocarbon.  The model needs the lake to have refilled after the Flood, which puts the radiation in under 4,400 years and makes the clock 3.3 times too long.  No young-earth author has published an age for the lake.  And no cichlid species was watched forming.

"The model is unfalsifiable."  Parts of it are not.  A founding pair predicts four or fewer deep sequences in any short window across a whole kind.  It predicts immune genes built as mosaics of four.  It predicts common variants near quarter frequencies.  Tracking predicted directed inheritance and failed.  The unfalsifiable part is the move Carter makes when he allows any amount of created variety in any number of founders.  The model is scientific to the extent that its authors refuse that move.

"Fast speciation from standing variation is just evolution, so creationists have conceded."  They have conceded that species change and split, and Wood dates the concession to 1941.⁴  Answers in Genesis calls the stickleback shift evolution in the textbook sense.⁷³  What they deny is that mutation and selection built the variation.  This article found no observed case that settles that either way, which is why the series measures it organism by organism in Benchmarking Evolution.

Issues and unanswered questions

1. The mutation-signature test on the old Eda, cichlid and human differences has not been run. It is the first thing to do. A 2021 study of the gray mouse lemur reports that across six primates the spectrum of differences between species was generally consistent with the spectrum of new mutations in families.⁷⁴ Whether any of its comparisons fall inside one kind was not checked.
2. No one has counted deep sequences per window across an animal kind. The dog family is the best sampled. The mouse family is the sharpest case: its founders would need about 17 percent difference between copies, and every murid in a short window should share at most four deep versions.
3. Nobody has priced the same between-species distance for the cat, horse or dog kinds, or asked whether two chromosome sets that different could work together in one animal.
4. Do the 90 dog and wolf versions of DLA-DRB1 reduce to mosaics of four?
5. Do species-rich kinds hold less variety per species than species-poor ones?
6. The gene-conversion rate used here is human. No rate for any wild animal was found.
7. Zangenberg 1995 and Arbanasić 2013 were read as abstracts only. The species that share wolf immune lineages are not known to this article.
8. Guliuzza's cavefish work was found only as a magazine summary. Whether the pigment change is inherited is not stated. The Pod Mrčaru lizards need the same test: eggs from each islet raised on the other diet.
9. How many fossil species each kind must also produce depends on where the Flood's rocks end, which creationists dispute. How far fossil species are oversplit is measured only for a few groups.
10. Whether the Nabugabo endemics are biological species has never been tested by breeding or genetics against their Victoria counterparts, and the lake's age rests on one indirect radiocarbon date.
11. Whether populations founded by a few animals in empty land escape the brake on sustained change in form that Uyeda's data show for established species.
12. Whether a Drift Lab panel on sorting from a founding pair belongs here is left for the editor.

Verdict

Stands.  Fast change and fast splitting run on variation already present.  This was watched in sticklebacks, dogs, flies and polyploid plants, and inferred in cichlids.  Creationist authors said so from the 1970s.  The mainstream reviews of 2008 and 2019 say the same.  Per lineage, the rate the average kind needs is about twice the fastest inferred, and the number of versions at immune genes is within reach of measured gene conversion.

Fails.  Continuous environmental tracking, at the one gene where inheritance was checked.  Fast mutation as a source of the differences, which the model's own authors rule out for nuclear DNA and which Wood must invoke for mitochondria with no mechanism.  Wood's requirement of 150 horse species in 400 years, which is about thirty times faster than any radiation known and which Wood himself says needs a mechanism not seen today.

Untested.  That the old variation was created, which for the mouse kind needs a founding pair about twice as variable as any animal measured.  That each kind passed through one pair about 4,400 years ago.  That designed mobile elements or directed mutation made new versions of genes on cue.  How much of each kind's variety was created and how much arose since.  Each has a test that could be run on public data, and none has been run by the people who propose them.

The model's mechanism is sound and its chronology is unsupported.  A skeptic is right to say that the clocks give millions of years and that no creationist has shown otherwise.  A creationist is right to say that the speed of speciation was never the problem it was made out to be.

Sources