The Human-Chimp Split and the Mutation Rate

"Although ubiquitous, this approach has an inherent circularity and is subject to possible error because it relies on the accuracy of the ages of fossils." — Kevin Langergraber and colleagues, on the old fossil-calibrated mutation rate, PNAS, 2012¹

"To say that God could have created the pattern is merely ad hoc." — Todd Charles Wood, young-earth creationist biologist, 2006²

Summary

Humans and chimpanzees differ at about 1.2 % of the DNA letters that can be lined up between them, roughly 35 million single-letter differences.³  How long would mutation take to make them?  The answer depends on the mutation rate, and for twenty years the field has had two rates that disagree by a factor of two.  This article works the arithmetic for both, then asks what a young-earth model has to say instead, and whether the evidence favors it.

1. The old rate was partly circular. The "phylogenetic" rate of about 1 × 10^-9 changes per DNA letter per year was set by dividing the human-chimp difference by an assumed divergence date of about 7 million years.¹ The "pedigree" rate is counted directly, by sequencing parents and children: 1.20 × 10^-8 per letter per generation in 78 Icelandic families.⁴ At human generation lengths of 25 to 29 years that is 0.41 to 0.48 × 10^-9 per year, about half the old rate.⁵
2. Half the rate means about twice the date. On the old rate, the leading genome model put the human-chimp split at 3.7 million years; on the measured rate, at about 6.⁶ Published estimates that use measured rates now run from 5.5-6.3 million years, from Yoo and colleagues in 2025 on complete ape genomes, through 6.5-9.3 from Moorjani and colleagues in 2016, to 7-13 from Langergraber and colleagues in 2012.¹⁷⁸
3. No single steady rate fits every fossil calibration, and the mainstream says so. The family rate fixes the human-chimp split but pushes the human-orangutan divergence to 34-46 million years, against a fossil limit under 18.⁹ The usual repair is a slowdown in the yearly rate as ape generations lengthened. Part of that repair is now measured: chimpanzee, gorilla and orangutan families mutate about 1.48 times faster per year than human families.¹⁰ This is a real weakness of the clock. It does not help a young earth: the youngest of these dates, 3.75 million years, is 500 times the roughly 7,500 years since creation on the Septuagint, the longest biblical chronology.
4. About a third of the genome does not follow the family tree. In 30 % of the genome, gorilla is closer to human or to chimpanzee than those two are to each other, split evenly between the two patterns, rarer near genes and rarer on the X chromosome.⁶⁶⁶ The 2025 complete genomes raise it to 39.5 %.⁸ On common descent this is "incomplete lineage sorting", and with the measured mutation rate it implies an ancestral population of tens of thousands, about 200,000 in the newest estimate. On separate creation the three genomes would have had to be made with this mosaic already in them.
5. Young-earth geneticists accept the measured differences and say most were created. Robert Carter writes that dating the ape lineages is only "uncovering the differences God built into their initial genomes."¹¹ The arithmetic leaves no other option. At measured rates, and even allowing for the patriarchs' great ages, mutation since creation adds about 7,000 to 9,000 changes to each lineage, depending on the chronology and the generation length. Both lineages together make 0.04 to 0.05 % of the 35 million differences. Mutation would have to run about 1,900 to 2,500 times faster to make them all.¹²
6. The "created" differences carry the fingerprints of mutation. A quarter of the human-chimp differences sit at CpG sites, which are 2 % of the DNA,³ and CpG sites are the hottest spots for new mutations in families.⁴ The Y chromosome differs most and the X least, in the ratio a male-biased mutation process predicts,³ and fathers pass on about four times as many new mutations as mothers.⁴ A designer could have made created differences look like this. No young-earth model predicts it, and Wood's warning about ad hoc appeals applies.
7. The waiting-time argument is the young-earth side's strongest point, and it proves less than is claimed. John Sanford's simulations find 84 million years to establish one specific two-letter DNA string in a hominin population, and 2 billion years for five letters.¹³ Mainstream mathematicians agree in substance: 216 million years for one prespecified pair of regulatory mutations in humans.¹⁴ But the wait shrinks in proportion to the number of acceptable targets. Inside 6 million years a two-letter string needs 15 equally good alternatives; a five-letter string needs 385.¹⁵ Nobody on either side has counted how many human-specific changes needed one exact multi-letter target.

Verdict.  The family-measured mutation rate is real, it roughly doubles the dates, and it leaves the clock needing a rate slowdown that is plausible and partly measured but not proven.  A young-earth model must say that more than 99.9 % of the single-letter differences were created.  The pattern of those differences is the pattern mutation and shared ancestry predict:  the CpG excess, the sex-chromosome ratio, and the mosaic of incomplete lineage sorting.  A created-differences model fits that pattern only by stipulation.  On this evidence the mutation reading is better supported.  The waiting-time problem is a real constraint on prespecified multi-letter changes, but it has not been shown to apply to the changes that actually separate the two species.

Part 1: Two clocks that disagree by a factor of two

A mutation rate is the chance that one DNA letter changes in one generation.  Turning it into a clock needs two more numbers:  how many letters differ between two species, and how many years a generation lasts.  Because both lineages mutate after they part, the time since their last shared ancestor is the difference divided by twice the yearly rate.⁵

The phylogenetic rate was set by the date it was later used to find

Before 2010 almost every human-chimp date used a rate of about 1 × 10^-9 per letter per year.  Langergraber and colleagues explain where it came from.  It was "derived from the observed DNA sequence difference of around 1.3% between the human and chimpanzee genomes ... and an assumed DNA sequence divergence between these species at 7 Ma," with a second calibration from human and macaque at an assumed 25 million years.¹  The arithmetic checks:  1.3 % divided by twice 7 million years is 0.93 × 10^-9.⁵

A date found with this rate is therefore partly the date that went into it.  Langergraber's own team calls the method circular.  The same paper also records the clock's best success.  In 1967 Vincent Sarich and Allan Wilson used immunological data to put the human-ape split at only about 5 million years, "some three to four times more recently than had been assumed on the basis of the fossil record."  DNA studies later put it at 4 to 6 million years again and again.¹

The pedigree rate is counted, not assumed

Since 2010, whole-genome sequencing of parents and children has let geneticists count new mutations directly.  Augustine Kong and colleagues at deCODE genetics sequenced 78 Icelandic families and found 4,933 new mutations, a rate of 1.20 × 10^-8 per letter per generation.⁴  The count rose with the father's age by about two mutations per year of his age.⁴  By 2012 four such family studies gave 0.97 to 1.36 × 10^-8 per generation.¹  The largest since, of 1,548 Icelandic families in 2017, gave 1.29 × 10^-8 per generation, or 0.43 × 10^-9 per year.¹⁶

The rate per generation is not yet a clock.  It has to be divided by the length of a generation, and that is not one number.

Every row is about half the phylogenetic rate.  Moorjani and colleagues put it plainly:  the family rate "is approximately twofold lower than the mutation rate inferred from the number of substitutions observed between primates."⁷  Aylwin Scally and Richard Durbin, reviewing the change in 2012, called the measured value "approximately half of that previously derived from fossil calibration."¹⁸  Generation length alone cannot close the gap.  They calculated that a generation of 10 to 15 years would be "needed for the de novo and phylogenetic rates to match," which "would be shorter than any other living great ape and less than half of the value measured in present-day humans."¹⁸

A third method gives a faster rate

Counting new single-letter mutations is not the only direct route.  James Sun and colleagues counted 2,058 new mutations at short repeated DNA segments in 85,289 Icelanders, then used the known behaviour of those segments to infer the single-letter rate.  They got 1.82 × 10^-8 per generation, 50 % above the family counts.¹⁹  The authors suggested the whole-genome studies might be missing some real mutations;  the whole-genome studies have since grown much larger and kept their lower value.¹⁶  The most complete count so far, published by David Porubsky and colleagues in 2025, read a four-generation family of 28 people with five sequencing technologies.  It covered 91.9 % of the 22 ordinary chromosomes, including the repeats and centromeres earlier studies could not read, and found 1.17 × 10^-8 per letter per generation.²⁰  New mutations were commoner in those repeats, but not common enough to raise the genome-wide rate.  So the gap is not explained by mutations hidden in hard-to-read DNA.  The disagreement between methods is itself one of the uncertainties in the dates below.

The other apes mutate faster per year

If the human rate per year had always been what it is now, the human, chimpanzee and gorilla lineages should show the same amount of change since they parted.  They nearly do.  Moorjani's team found substitution rates "∼2% higher in chimpanzees and ∼7% higher in the gorilla than in humans."⁷  Yet Søren Besenbacher and colleagues, sequencing chimpanzee, gorilla and orangutan families, found these apes now mutate about 1.48 times faster per year than humans.¹⁰  The two findings fit only if the human slowdown is recent, and that is how Besenbacher reads it:  "an appreciable slowdown in the yearly mutation rate in the human lineage that is likely to be recent as genome comparisons almost adhere to a molecular clock."  The favoured cause is life history.  Most new mutations are counted per generation, so a species that breeds younger collects more of them per year, and Scally and colleagues note that ape body sizes and generation times grew over this period.⁶  Generation length alone is not enough:  as shown above, it would need 10- to 15-year generations.  Guy Amster and Guy Sella's 2016 model adds "the age of onset of male puberty and rates of spermatogenesis," since most new mutations arise in fathers.  With those, they report, it "largely bridges the gap reported between estimates of split times based on fossil and molecular evidence," and the split can be as recent as 6.6 million years.²¹  The price is inputs nobody can measure:  how old ancestral apes were at puberty and at fatherhood millions of years ago.

Part 2: From a difference to a date

Two clocks run inside every comparison

A human and a chimpanzee DNA sequence did not start to differ on the day the two populations parted.  The ancestral population already carried variation, so two gene copies picked from it had usually split long before.  Langergraber and colleagues state it this way:  the differences seen today are "the sum of two values," those that built up after the two lineages stopped interbreeding and "the sequence differences that correspond to the diversity in the common ancestor of both lineages."¹  The 2005 chimpanzee genome paper splits the time the same way, into the time since speciation and the coalescence time in the ancestral population, which it put on the order of 1 to 2 million years.³

So a date needs three things:  the difference, the yearly rate, and the size of the ancestral population.  For two gene copies in a population of N breeding individuals, the average wait back to their common ancestor is 2N generations.  The split time is the total divergence time minus that wait.⁵

The arithmetic, at each rate

Divergence of 1.23 %, the 2005 consortium's figure,³ gives these times since the last shared ancestor of the DNA, before any allowance for the ancestral population:

All figures are in millions of years, with a 24.5-year generation in the ancestor.⁵  Two things follow.

First, the old rate leaves almost no room for an ancestral population.  At 1 × 10^-9 per year, a 6-million-year split allows an ancestral population of only about 3,000, and a 7-million-year split is impossible.  The measured rates need a large ancestral population for a 6 to 7 million year split:  about 100,000 to 180,000.  The 2025 complete-genome study estimated 198,000.⁸

Second, rate and population size trade against each other.  A slower rate makes the total divergence older, and a larger ancestral population takes more of that time away from the split.  Much of the spread between published dates comes from different choices of these two inputs.

What the published studies found

Langergraber's team recalculated older published splits with the family rates.  Scally's 3.69 million years became 6.78 to 11.03.  Others moved from 4.2-4.5 to between 7.8 and 13.5.¹

The orangutan problem

The family rate fixes one fossil problem and creates another.  On the old rate, the human-chimp split fell after fossils such as Sahelanthropus, dated 6.8 to 7.2 million years, that many paleontologists place on the human side.¹⁹  The slower rate moves the split back past them.  Whether Sahelanthropus stood on the human line at all is still argued, which makes it a soft calibration either way;  that dispute is worked in Sahelanthropus, Orrorin, and Ardipithecus.  But Sun's team found that the same slow rate puts the human-orangutan divergence at 34 to 46 million years, "so much older than the upper bound from the fossil record of <18 Mya on human-orangutan speciation that the date is implausible."⁹

Scally's team proposed the repair now usually accepted:  a rate near 1 × 10^-9 in the early great apes, slowing as body size and generation length grew in every descendant line.  They admitted the cost:  Sahelanthropus and an early gorilla fossil "can be accommodated ... only if most of the decrease occurred early in great ape evolution."⁶  Sun's team was blunter.  The slowdown "would need to have been coincidental in both lineages," and "extraordinarily dramatic:  about 3-fold in both lineages."⁹  Besenbacher's ape families have since measured part of that difference directly, and with their faster ape rates the dates come out "much more in line with the fossil record and the biogeography."¹⁰

Scally summed up the state of the field in one sentence:  "the inability to observe past mutation rates means that the timing of events from genetic data remains uncertain."⁶  Geneticists said as much in public.  At a 2015 meeting in Leipzig on the human mutation rate, David Reich of Harvard Medical School told Nature:  "The fact that the clock is so uncertain is very problematic for us," and "It means that the dates we get out of genetics are really quite embarrassingly bad and uncertain."  The same report noted that the slow clock "gave nonsensical results when extended further back in time," such as an ape-monkey ancestor that "could have encountered dinosaurs," and that many authors were "presenting multiple dates" to hedge.  Reich still advised using the slower rate, with a warning:  "This is not the speed of light.  This is not physics."²²  Young-earth critics quote Reich's words, and on the uncertainty they are right.²³  A skeptic should grant the point.  So should a creationist reading it as support for a young earth, because every candidate date in this dispute lies between 3.75 and 13 million years.  On the Masoretic text's roughly 6,000 years since creation that is 600 to 2,200 times too old;  on the Septuagint's roughly 7,500, 500 to 1,700 times.  The disagreement is a factor of about three and a half.  The gap to a young earth is a factor of a thousand.

Part 3: Incomplete lineage sorting, a third of the genome off the family tree

What it is

Take any stretch of DNA and ask which two of human, chimpanzee and gorilla match best.  Most of the time the answer is human and chimpanzee.  But in 30 % of the genome, "gorilla is closer to human or chimpanzee than the latter are to each other."⁶  The complete genomes of 2025 put it at 39.5 % of the autosomes and 24 % of the X chromosome.⁸

Common descent explains this without strain.  The population ancestral to humans and chimpanzees carried many variants, some of which it had inherited from the older population that also gave rise to gorillas.  If the time between the gorilla split and the chimpanzee split was short compared with the size of that population, many variants were still unsorted when humans and chimpanzees parted.  Each lineage then kept one version or the other more or less at random.  Where the human line kept the version the gorilla line also kept, human and gorilla match.

The proportion follows a simple rule.  Two gene copies fail to find their common ancestor in the population between the two splits with a chance of e^-t/2N, where t is the time between the splits in generations and N the population size.  When they fail, two times in three the resulting tree disagrees with the species tree.²⁴  A 30 % disagreement means the time between splits was only about 0.8 times twice the population size.  If that population had held 10,000 individuals and lasted 4 million years, the disagreement would be 0.007 %, not 30 %.

ILS fixes only that ratio.  It does not fix years, because a population of 100,000 lasting 3.5 million years gives the same share as one of 50,000 lasting 1.7 million.  The mutation rate sets the years, which is why the rate dispute and the population-size estimates move together.  With the measured rates, published estimates of the ancestral population run from about 50,000 to about 200,000.⁷⁸⁵

It passes three checks a random mosaic would not

ILS is not a fudge factor added afterward.  It predicts specific patterns, and the genomes show them.

1. The two odd trees should be equally common. Chance sorting has no favourite. Scally's team found "no significant difference between the number sorting as ((H,G),C) and ((C,G),H)."⁶ Interbreeding between two of the species after they parted would have favoured one of the two.
2. It should be rarer where natural selection shrinks the effective population. Selection on genes removes variation from the DNA around them. ILS drops to 22 % inside coding exons, and "the suppression of ILS extends out to several hundred kbp from coding genes, evident even in raw site patterns before any model inference."⁶
3. It should be rarer on the X chromosome. A population carries three X chromosomes for every four copies of each other chromosome, and selection acts on the single X in males. Scally's team found less ILS on X;⁶ the complete genomes give 24 % on X against 39.5 % on the other chromosomes.⁸

The same population variation also accounts for a large share of the human-chimp difference itself.  On the mainstream models, between a fifth and three fifths of the 1.2 % was already present as variation before the two lineages parted.⁵  Nathaniel Jeanson asked evolutionists about such "preexisting differences":  "When can preexisting differences be invoked, and when should they not be invoked?  What testable predictions does the hypothesis of preexisting differences make?"²⁵  It is a fair question, and ILS is the answer.  The preexisting variation of the mainstream model must show up as gene trees that disagree in a known share, evenly between the two alternatives, less near genes and less on X.  It does.

What a separate-creation model would need

On a young-earth model humans, chimpanzees and gorillas were created as separate kinds, with no shared ancestral population.  The pattern then has to have been built into the created genomes:  in about a third of the DNA, human made to match gorilla against chimpanzee or chimpanzee to match gorilla against human, in equal measure, less often near genes and on the X chromosome.

Jeffrey Tomkins and Jerry Bergman, writing in the Journal of Creation in 2013, take ILS as evidence against common descent.  They write that ILS "across clearly different and unrelated kinds of organisms represents a mosaic of DNA sequence patterns that cannot be explained by common ancestry."²⁶  On the gorilla genome they say that "no consensus path of common ancestry between humans and various apes exists, and no coherent model of primate evolution could be achieved."²⁶  That misreads the paper they quote.  Scally's team fit the 30 % inside one model with one set of speciation times and population sizes, and wrote that "whole genome comparisons can be strongly conclusive about the ordering of speciation events."⁶  What the genome left uncertain was the timing, not the tree.

No young-earth writer found in this research has offered a created-kinds model that predicts the size of the ILS share, its even split, or its drop near genes and on the X.  Jeanson's Replacing Darwin, the fullest young-earth treatment of nuclear DNA differences between species, does not discuss it.²⁷  A designer could have made the three genomes this way.  But as Todd Wood wrote of the human-chimp pattern in general, "To say that God could have created the pattern is merely ad hoc."²

Part 4: The young-earth position, in its own words, and its price

Young-earth geneticists accept the measured differences

None of the young-earth sources examined here disputes that humans and chimpanzees differ at tens of millions of DNA letters.  Their disagreement is about where the differences came from.

Nathaniel Jeanson, whose 2017 book Replacing Darwin gives the fullest young-earth treatment, accepts the measured rates and the measured differences and runs the evolutionary prediction himself.  "Traditionally, evolutionists have put the split time between these species at 4.5 to 7 million years ago."  At the measured human and chimpanzee rates that predicts "only around 8.4 million" to "only around 13.1 million" differences, while "over 26 million single base pair differences separate the human and chimpanzee nuclear DNA sequences."²⁵  He then notes that the date moved to fit:  "The human-chimpanzee divergence time was revised in light of the mutation rates in these species."²⁵

That observation is correct, and it is the same gap Langergraber and Scally addressed in 2012.  Two corrections are needed to his arithmetic.  First, his prediction counts only the time since the split and leaves out the variation already present in the ancestral population, which the 2005 chimpanzee genome paper had put at 1 to 2 million years' worth of divergence.³  Second, the "11 to 17 million years" he reports as the revised split²⁵ is, in the chimpanzee family study it matches, "an average time to the most common ancestor," the total divergence time rather than the split.²⁸  A total divergence near 13 million years, with the split several million years younger, is what Part 2's arithmetic gives at that rate.  Jeanson's second argument, that the fast family rate of mitochondrial DNA predicts far more human-chimp mitochondrial differences than the 1,483 observed, turns on why short-term and long-term mitochondrial rates differ.  That question belongs to the Mitochondrial Eve article.

His own answer is creation:  "A creationist would respond by saying that these differences were created."²⁵  He sums up:  "it appears that mtDNA differences are due to a constant rate of mutation over 6,000 years, whereas the majority — if not vast majority — of nuclear DNA differences are due to preexisting DNA differences that were placed in the ancestors of modern species 6,000 years ago."²⁵

Robert Carter of Creation Ministries International takes the same view of the ape dates:  "As far as timing the rise of the great ape lineages, all they are doing is uncovering the differences God built into their initial genomes.  One of the basic premises of creation thought is that God designed hierarchically ... so some pairs of created groups are necessarily more similar than others and uncovering those differences does nothing to address the creation-evolution debate."¹¹

Todd Wood, the young-earth biologist who has written most carefully about the chimpanzee genome, is more cautious.  He accepts the consortium's figures, including its estimate that the fixed difference may be "as low as 1.06%."²  He floats, and does not endorse, the idea that God "created humans and chimpanzees with identical genomes," calling it "an intriguing possibility" that known biochemical differences "may well rule ... out."²  And he names the problem young-earth genetics has not solved:  "Because many differences between the human and chimpanzee genomes may have been created rather than arising during post-creation history, we are left with the challenge of distinguishing created differences from differences that arose during the history of the two baramins."²  His closing judgment:  "The genome revolution, exciting though it is, is not an obvious victory for creationism."²

John Sanford frames the differences as information the human line would have had to gain:  "Chimp and human genomes are minimally 5 % divergent [1], representing about 150 million nucleotide differences."¹³  That figure counts every letter inside insertions and deletions as a separate difference.  Wood made the correction in 2006 against an earlier creationist argument:  "An insertion or deletion of 1000 nucleotides is only one mutational event, even though the total difference is 1000 nucleotides."²  The single-letter differences number about 35 million;  the insertion and deletion events about 5 million.³

The price: more than 99.9 % created

The arithmetic leaves a young-earth model no middle ground.  Take the measured human rate, about 35 new single-letter changes per copy of the genome per generation.  Give the patriarchs of Genesis 5 and 11 the extra mutations that older fathers pass on today, about two more per child for each year of the father's age, run in a straight line up to Noah's 502 years.  From Adam to Abraham that is 19 births in 1,948 years.  Then add about 145 to 168 ordinary generations to the present.⁴¹²

The Septuagint's longer ages before Abraham raise the patriarchs' share from about 2,000 to about 3,400, and the total to about 8,400 to 9,200 per lineage, or 0.05 % of the differences.  To make all the differences by mutation in this time, the rate would have to run about 1,900 to 2,500 times faster than it is measured in families today, in both species, on every line.  That would be about 130,000 to 175,000 new mutations in every child, against about 60 to 70 today.

Two young-earth geneticists have floated a steeper effect of a father's age, and it does not change the answer.  Robert Carter and Matthew Powell, extrapolating from a 2000 review of human mutation by the geneticist James Crow, wrote that "a man 500 years of age would donate approximately 10,000 mutations to a child."  On Kong's straight line, they note, Noah "would only contribute slightly more than 1,000 mutations."²⁹  Giving 10,000 new mutations to every one of the ten pre-Flood births, not to Noah alone, still accounts for only 0.6 % of the differences.¹²

So Wood's "identical genomes" possibility is ruled out by the numbers as well as by the biochemistry he mentions, and Jeanson and Carter are right that, on a young earth, almost all of the differences must have been there from the start.  The young-earth claim to test is therefore not "mutation was faster."  It is "more than 99.9 % of the single-letter differences between humans and chimpanzees were created."  Part 5 tests it.

Part 5: Created differences or mutations? The fingerprints

Mutation is not random in where it strikes or what it does.  It has habits, measured directly in families.  If the human-chimp differences were made by mutation, they should share those habits.  If they were created, nothing requires them to.  Three habits can be checked against published numbers.

The CpG hotspot

Where a C is followed by a G in DNA, the C usually carries a methyl tag, and a methylated C decays to T far more often than any other letter changes.  Kong's family study found transitions at CpG sites occurring at 18.2 times the rate at other sites, "ostensibly because spontaneous oxidative deamination of methylated cytosines leads to an increase in transition mutations."⁴

The human-chimp differences show the same hotspot.  CpG sites "account for 25.2% of all substitutions while constituting only 2.1% of all aligned bases," a twelvefold excess and about 9 million of the 35 million differences.³⁵  Two finer details match too.  Changes at CpG sites behave more like a clock in years than other changes do, which Moorjani's team attributes to "their nonreplicative origin":  they come from chemical decay over time rather than from copying errors.⁷  And the 2005 consortium found the male excess at CpG sites much smaller than elsewhere, about 2 against 3 to 6, which it read as "roughly equal rates of CpG to TpG transitions in the male and female germ line."³  Decay that runs with time should hit eggs and sperm alike;  copying errors pile up in sperm, which go through far more cell divisions.

The sex chromosomes

Fathers pass on more new mutations than mothers.  In Kong's families the ratio was 3.9 to 1;⁴ in a chimpanzee family studied by Oliver Venn and colleagues, "seven to eight times."²⁸  This bias has a predictable effect on long-term change.  The Y chromosome is always in a male, the X spends two thirds of its time in females, and the other chromosomes half.  So with a male bias the Y should change fastest and the X slowest.

The human-chimp differences do exactly that:  "a mean divergence of 1.9% for chromosome Y and 0.94% for chromosome X," against 1.23 % overall.³

Both observed ratios fall inside the range the family-measured biases predict.³⁰  The match is not perfect.  The X is somewhat less different than the human bias alone predicts, which Scally's team put down to the smaller ancestral population of X chromosomes, and after correcting for it they estimated a male bias of 2.3.⁶

Regional variation

Mutation rates vary along the chromosomes, and so do the human-chimp differences.  The 2005 consortium concluded that "the large-scale variation in the human–chimpanzee divergence rate primarily reflects regional variation in mutation rate."³  The direct test came later, by comparing where new mutations land in families with where the human-chimp differences are densest.  Thomas Smith, Peter Arndt and Adam Eyre-Walker found the two correlated in all three large family datasets, "as expected," but more weakly than mutation alone predicts:  0.16 to 0.31, against 0.30 to 0.68 expected.³¹  They traced part of the gap to recombination, which also biases which variants become fixed.  This check counts for mutation, but less strongly than the other two.

What these fingerprints do and do not show

They show that the human-chimp differences have the statistical character of mutations, in three independent respects, each measured in families without assuming any ancestry.  They do not by themselves prove common descent.  A model in which separately created kinds then diverged by mutation would also produce them, but Part 4 showed a young-earth timescale leaves room for only 0.05 % of the differences to arise that way.

So the young-earth model must say that God created the differences with the fingerprints of mutation already in them:  a twelvefold CpG excess, a Y-to-X ratio matching the male bias of fathers' sperm, and the regional pattern of mutation rates.  That is possible for a Creator.  It is also exactly what Wood called "merely ad hoc,"² and it gives up the one test that could separate created from mutated differences, the test Wood said young-earth genetics most needed.²  No young-earth source examined here addresses the CpG excess or the sex-chromosome ratio.²⁷

What fits the created-differences view better

The created-differences view is not without support.  Four findings sit awkwardly with a simple mutational clock.

1. The yearly rate is not constant. As Part 2 showed, no single steady rate fits both the human-chimp and the human-orangutan fossils. The repair, a slowdown as apes reached puberty and fatherhood later, is reasonable and partly measured in living apes, but its timing and the ancestral ages it needs are inferred, not observed.
2. The direct methods disagree. Repeat-segment counting gives a rate 50 % higher than whole-genome family counting.¹⁹
3. The regional match is weaker than mutation alone predicts.³¹
4. Single-letter changes are not the whole difference. Insertions, deletions and rearranged segments cover far more DNA than single-letter changes do, and the 2025 complete genomes found 12.5 to 27.3 % of each ape genome without a simple one-to-one match. Those differences are the subject of How Similar Are Human and Chimp DNA?. This article prices only the single-letter changes.

None of these four is evidence that the differences were created.  Each is a place where the mutational account has a free parameter or an unexplained residue.

Part 6: The waiting-time problem on the human line

The questions so far concern neutral changes, which make up most of the human-chimp difference and need no selection.  The waiting-time argument asks something else:  how long a hominin population must wait for a specific useful change that needs several DNA letters at once.

The young-earth case, and the mainstream agreement

John Sanford and colleagues modified the simulation program Mendel's Accountant and "simulated a classic pre-human hominin population of at least 10,000 individuals, with a generation time of 20 years, and with very strong selection."  Their result:  "To establish a string of two nucleotides required on average 84 million years.  To establish a string of five nucleotides required on average 2 billion years."¹³  Their full table gives 1.53 million years for one specified letter, 84.1 million for two, 2.31 billion for five and 18.5 billion for eight.¹³  Sanford concludes that "Routine establishment of specific beneficial strings of two or more nucleotides becomes very problematic."¹³

Mainstream mathematics agrees on the core.  Rick Durrett and Deena Schmidt, writing in Genetics in 2008, worked out how long a human-sized population waits for one prespecified pair of mutations that switches off one gene-control site and creates another:  "Multiplying by 25 years per generation gives 216 million years."  The same paper notes that a new control site needing only one mutation appears "in an average of 60,000 years."¹⁴

The price

The human lineage has had 5.5 to 13 million years since the split, on the dates of Part 2.  Against that window:

The last two columns use a property both sides accept:  if any of k different strings or places would do the job equally well, the expected wait falls by about k.¹⁵  Sanford says so himself.  He lists the ways a real target could be easier than his, including that "the string should be able to arise anywhere," and concludes, "In all these special circumstances the waiting time will certainly be shorter than this study indicates."¹³  He also reports a published model in which a five-letter string allowed anywhere among 20,000 human gene promoters takes only about 7,000 years.  He rejects it on biological grounds:  "biologically it is simply not reasonable to assume that a specific short string can be, by itself, beneficial almost anywhere in the genome, with minimal contextual constraint."³²¹³

What the argument shows

The arithmetic is not in dispute.  A change that needs two or more particular letters at one particular place, with no benefit until all are present, is very unlikely on the human line in 6 or even 13 million years.  A single specified letter is not limited:  Sanford's own model gives it 1.5 million years.

What is not known is how many human-specific traits needed changes of the hard kind.  If most needed one of hundreds of possible strings, or could be built one useful letter at a time, the waiting time vanishes.  If even a few needed one exact multi-letter string with useless intermediates, it bites.  No published study on either side has gone through the human-specific changes and counted.  Until someone does, the waiting-time argument establishes a constraint, not a refutation.

The separate question of how many beneficial changes selection could fix in the time, Haldane's dilemma, and the full technical dispute over Sanford's simulations are in Haldane's Dilemma and the Waiting-Time Problem.

Objections and responses

"Molecular clocks assume common ancestry, so any date they give is circular."  Partly true, and the mainstream says so about the old rate:  Langergraber's team called fossil calibration "an inherent circularity."¹  The family rate is different.  It is counted in living parents and children and assumes nothing about ancestry.  Turning it into a split date does assume the two lineages shared an ancestor, so a date cannot by itself prove shared ancestry.  That is why this article tests the pattern of the differences in Parts 3 and 5, not only their number.

"The rates disagree by a factor of two or more, so the clock is worthless."  The spread among published human-chimp splits is about three-and-a-half-fold, from 3.75 to 13 million years.  The gap between any of them and a young earth is about a thousandfold.  A clock that cannot tell 4 million years from 13 can still tell either from 7,500.

"Rates measured in families run far faster than fossil-calibrated rates, so the long ages are wrong."  Jeffrey Tomkins and Jerry Bergman, writing in the Journal of Creation in 2015, call counting in families "the most empirical way to measure genetic change" and say these counts "greatly exceed the more speculative and spurious rates achieved from paleontology."²³  That is true of the fast-changing control region of mitochondrial DNA, as the Mitochondrial Eve article shows.  For the nuclear DNA that holds the 35 million differences, it runs the other way.  The family rate is about half the fossil rate,¹⁸ so the most empirical rate makes the human-chimp split older, not younger.  Tomkins and Bergman also argue that "pervasive genome-wide biochemical functionality" undercuts the neutral theory on which clocks rest.²³  If most of the differing letters did have functions, the dates would grow longer still.  Changes with no effect pile up at the rate mutations arise.²¹  Changes that damage a function are mostly weeded out, so working DNA collects differences more slowly than mutation supplies them, and more time is needed to reach the same count.

"The evolutionists simply moved the date to fit the new rate."  Jeanson is right that the date was "revised in light of the mutation rates."²⁵  Revising a date when a better measurement arrives is how measurement works.  The real cost is elsewhere:  the revision broke the orangutan calibration, and the repair, a slowdown in the yearly rate, is a free parameter that has been measured only in living apes.⁹¹⁰

"Young-earth geneticists trust family-measured rates for Mitochondrial Eve.  Consistency demands the same rate here."  They do, and the consistency holds.  The same family rate that young-earth writers use for human mitochondrial DNA gives at most about 0.05 % of the human-chimp differences in 6,000 to 7,500 years.  Jeanson and Carter accept this, which is why they say the differences were created.  The tension is not inconsistency.  It is that the created differences look like mutations.

"God would naturally use the same chemistry and the same design rules for both species, so created differences would look like mutations."  Shared chemistry explains shared genes.  It does not explain why the differences pile up at the one site that methyl-C decay attacks, or why the Y differs twice as much as the X in the ratio set by fathers' sperm.  Those are signatures of a process acting over time, not of a design rule.  A Creator could have reproduced them, but the explanation then predicts nothing that was not already observed, which is Wood's objection to ad hoc design.²

"Incomplete lineage sorting shows the tree of life is false."  ILS needs no transfer of DNA between species after they parted;  Tomkins and Bergman assume it does.²⁶  It needs only that variants present before the split were kept unevenly.  It predicts an even split between the two odd trees and a drop near genes and on the X, and the genomes show both.⁶

"Humans and chimpanzees differ by 5 %, 150 million letters, far too many for mutation."  The 5 % counts every letter inside an insertion or deletion as a separate change.  Wood corrected the same mistake in 2006.²  The single-letter differences number about 35 million, the insertion and deletion events about 5 million.³  On the mainstream model, the single-letter count is what the measured rate predicts once the ancestral population is counted, as Part 2 shows.

"The waiting-time problem proves humans could not have evolved from apes."  It proves that changes needing a prespecified multi-letter string at one place could not arise in the time.  Whether human evolution needed many such changes is not established by Sanford, who concedes the waits shorten whenever the string could arise in more places or more forms.¹³

From the skeptic's side:  "The young-earth model makes no testable prediction here at all."  It can.  Wood posed the test:  distinguish created differences from those that "arose during the history of the two baramins."²  A created-differences model that predicted, in advance, a spectrum of differences unlike mutation's, and then found it, would be strong evidence.  The spectrum found so far is mutation's.

Issues and unanswered questions

1. Besenbacher's faster ape rates rest on ten families: seven chimpanzee, two gorilla and one orangutan. The orangutan interval, 0.98 to 1.85 times the human rate, includes no difference at all.¹⁰ Method matters too. When five laboratories analysed the same family of rhesus macaques in 2022, their rates ran from 0.46 to 0.85 × 10^-8 per letter per generation, "almost a twofold variation" from processing choices alone.³³ Besenbacher's team guarded against this by running the apes through "essentially the same" pipeline it had used for humans, and by checking that about half of the new mutations found in a young ape passed on to its own offspring, as real mutations should.¹⁰ The slowdown that rescues the orangutan calibration is measured, but thinly.
2. The two direct methods still disagree. Repeat-segment counting gives 1.82 × 10^-8 per generation and whole-genome counting about 1.2 × 10^-8. Sun's team suggested the genome counts miss real mutations;¹⁹ the genome counts have since grown much larger without rising, and the 2025 four-generation family, read into the repeats and centromeres, still gave 1.17 × 10^-8.²⁰ No paper found here reconciles them.
3. The transition-to-transversion ratio of human-chimp differences was not found stated in any source read, so it could not be compared with the 67.8 % transitions seen in families.⁴
4. The regional correlation is half of what mutation alone predicts. Smith and colleagues attribute the rest partly to recombination and biased gene conversion.³¹ A young-earth reader could press this point harder than the mainstream does.
5. No created-kinds model of ILS exists. A model that produced the 30 to 40 % mosaic, its even split, and its drop near genes and on the X from designed starting genomes would be a real contribution. None was found.
6. Nobody has counted the hard changes. The waiting-time dispute turns on how many human-specific traits needed one exact multi-letter string. Neither side has done the count.
7. Jeanson's "11 to 17 million years" is cited in his book to "the authors of the chimpanzee genome paper." The figure matches Venn and colleagues' 2014 chimpanzee family study, which calls it an average time to the common ancestor. The 2005 chimpanzee genome paper does not contain it.²⁸²⁵

Verdict

The family-measured mutation rate is the best-measured number in this debate, and it is about half the old fossil-calibrated rate.  With it, the human-chimp split moves from about 4 million years to between 5.5 and 13 million, depending on the generation length and the ancestral population assumed.  The revision fixes one fossil conflict and creates another, which mainstream geneticists repair with a slowdown in the yearly rate that is plausible and partly measured.  That is a real weakness, and young-earth critics are right to press it.  It moves the dates by a factor of about three, never toward thousands of years.

A young-earth model cannot get the differences from mutation.  At measured rates, even allowing for the patriarchs' great ages, mutation since creation makes 0.04 to 0.05 % of them.  Jeanson and Carter accept this and say the differences were created.  That claim is coherent, but it meets three findings that it does not predict:  the differences sit at mutation's CpG hotspot, the sex chromosomes differ in the ratio set by fathers' male-biased mutation, and a third of the genome shows the gorilla mosaic that a large shared ancestral population leaves behind, in the proportions and places that theory expects.  The mutational reading predicts all three.  The created reading can only accommodate them.

The waiting-time argument is the young-earth side's strongest point.  It shows that changes needing exact multi-letter strings are very unlikely on the human line.  It has not shown that the human line needed many such changes.

On the evidence available, the human-chimp differences are better explained as mutations accumulated over millions of years than as created differences.  The young-earth case here rests on Wood's unmet challenge, a way to tell created differences from mutated ones, and on a count of waiting-time targets that nobody has made.

Sources