FitCoal and the Bottleneck That Never Happened

When a computer model is given deep time, it returns deep time. That is not a discovery. That is a circle.

In 2023 a team of geneticists published a striking claim in Science. Using a new method they named FitCoal, the fast infinitesimal time coalescent process, they announced that human ancestors nearly vanished between about 930,000 and 813,000 years ago. The breeding population, they said, fell to roughly 1,280 individuals and stayed that small for about 117,000 years. Headlines followed. Humanity had brushed extinction long before “modern humans” were supposed to exist. The result was offered as a window into our ancestry and even as a possible moment of speciation.

It sounds like history recovered from the genome. It is not.

FitCoal does not read a date stamped in DNA. It takes the pattern of how often genetic variants appear in living people, the site frequency spectrum, and asks which imagined history of population size would produce that pattern. To turn frequencies into years and into “breeding individuals,” the method must be fed a list of assumptions. Those assumptions do the real work. Change them, and the 900,000-year bottleneck disappears. Keep them, and the model will keep writing Pleistocene novels out of post-Flood families.

We have seen this pattern before in our other articles on genetic clocks. The data are real. The calendar is imported.

What FitCoal Is Doing

Think of every person alive as carrying copies of DNA that meet, if you walk backward through the family tree, in shared ancestors. Population geneticists call those meetings coalescences. In a small population, lineages meet sooner. In a large population, they wander longer before they meet. The mix of rare and common variants in a sample is supposed to remember how often those meetings happened.

FitCoal slices time into millions of tiny steps, assumes the population size is constant inside each step, calculates how long the branches of that genealogical forest should be, and hunts for the size history that best matches the observed spectrum of variants. On paper it is an elegant piece of mathematics. In practice it is only as trustworthy as the world it is told to inhabit.

That world is not the world of Genesis.

The Assumptions That Write the Story

A mutation rate borrowed from deep time. The published analyses used a rate of about 1.2×10−81.2 \times 10^{-8}1.2×10−8 changes per DNA letter per generation and a generation length of 24 years. Those two numbers convert “how many generations” into “how many years.” They are not measured off the FitCoal genomes. They are inserted. If the real per-generation rate is faster, every date shrinks. If generations are longer or shorter, the calendar slides again. The method cannot discover 930,000 years. It can only report the years that its chosen clock allows.

This is the same circular habit we have already documented. Phylogenetic rates are slowed to match an assumed fossil timescale, then used to “prove” that timescale. Pedigree studies, which count new mutations from parent to child, tell a different story. When the clock is the one actually observed in families, the years collapse toward the history Scripture already gave us.

A well-mixed population that never existed. FitCoal treats the ancestry of a sample as if it were one panmictic pool, or as if each population can be read that way on its own. Real human history is families, tribes, languages after Babel, and later nations. Structure, isolation, and later mixing all change the frequency spectrum. A model that is not allowed to see that structure will often invent a crash in population size instead. The “bottleneck” can be the shadow of divided families, not a 117,000-year near-extinction.

Kingman’s coalescent and a long, smooth past. Lineages are assumed to meet two at a time across tens of thousands of generations. The history is a stack of quiet size changes, not a year-long global Flood, eight survivors, rapid growth, and then a sudden scattering of peoples. A tool built for that slow geometry cannot recognize the event Genesis 6 through 11 describes. It will smear a sharp, recent reset into a long, ancient dip, or it will push the dip so far back that it no longer means anything biblical at all.

Neutrality and independent sites. FitCoal prefers non-coding sequence and often drops high-frequency variants to reduce the effects of selection. That helps. It does not erase linked selection, mutation-rate differences along the genome, or the simple fact that many sites are not independent. Those distortions move the spectrum without any change in the number of people who were alive.

Effective size treated as a headcount. The number 1,280 is an effective size, NeN_eNe​, not a census of husbands and wives. Unequal family sizes, overlapping generations, and the way reproduction actually works make NeN_eNe​ much smaller than the number of human beings. Calling that figure “near extinction” oversells the parameter.

All differences treated as mutations piled up after a single evolutionary origin. Created variation in Adam and Eve, the second bottleneck of Noah’s household, and the designed diversity within the human kind are invisible to the model. Everything it sees it must explain as copying errors across vast time. That is not an inference from the data. That is the theory deciding in advance what the data are allowed to be.

Where the Assumptions Defeat the Claim

If the clock is assumed, the date is assumed. FitCoal’s central result is therefore not an independent testimony against Scripture. It is Scripture’s timeline being excluded at the first step.

If the population is assumed to be well mixed for hundreds of thousands of years, a structured post-Babel world will be misread. The method then “discovers” a crash that the authors themselves struggle to find cleanly in non-African samples, even though an event at 900,000 years ago should sit in the shared ancestry of every nation on earth. That inconsistency is the model colliding with its own premise.

If many different histories can produce similar frequency spectra, a flexible search that is allowed to add more and more time slices will prefer a dramatic dip. Other research groups have now said so in print. Simulations that resemble human data can trick FitCoal into drawing a sharp bottleneck that was never in the true history. Alternative models can fit the same living spectra as well or better without any 900,000-year collapse. When the claim fails under neighboring methods, the claim was never as solid as the press release.

If NeN_eNe​ is not a headcount, the extinction story is rhetoric. And if the only events the algorithm is allowed to propose are slow size changes in a Kingman process, then the Flood, the most important demographic event after creation, is ruled out before the first likelihood is calculated. A method that cannot see Noah will never confirm Noah. That is not a failure of Genesis. That is a failure of the instrument.

What Scripture Already Told Us

The genealogies of Genesis 5 and 11, read as they are written, place the creation of Adam near 4004 BC and the Flood near 2348 BC. That is about 6,000 years of human history and about 4,500 years since eight people walked off the Ark. “Of them was the whole earth overspread” (Genesis 9:19). A few generations later the Lord scattered the families at Babel (Genesis 11:8–9). Those are the bottlenecks that matter: a created pair, then a household of eight, then rapid multiplication and dispersion.

A 117,000-year trough in the early Middle Pleistocene is not a translation of that history into another language. It is a different religion of origins. The duration is wrong. The date is wrong. The cast of characters is wrong. No amount of smoothing the SFS will turn 1,280 mid-Pleistocene hominins into Noah, his wife, his three sons, and their wives.

What Measured Mutation Rates Actually Produce

When we begin with a Pleistocene clock. we also begin with mutations counted in families. On the Y chromosome, the chromosome sons inherit from their fathers, high-coverage pedigree studies yield about two to three new mutations per generation in the sequenced regions he uses. When that observed rate is applied to published global Y trees, the most recent common ancestor of living men falls about 4,500 years ago. That is the Flood generation. It is not 250,000 years ago. Mitochondrial DNA, passed from mothers, likewise fits a history measured in thousands of years, not hundreds of thousands, when the pedigree rate is used rather than a phylogenetic rate slowed to match fossils.

The shape of that Y tree, once it is allowed to be rooted where Scripture roots it, tracks the Table of Nations. Three deep trunks answer to Shem, Ham, and Japheth. Later branches follow the movements of their descendants. That is what a real post-Flood genealogy should look like. It is also why actual genetic data and FitCoal’s result cannot both be histories of the same world. One clock is taken from father and son. The other clock is taken from an assumed age of rocks.

Use actual measured rates inside an autosomal method like FitCoal and the calendar would not land at 930,000 years. The same pile of variants, counted at the speed families actually mutate, belongs in the few thousand years since Noah, with a large share of the differences already present as created and post-Flood family variation. The algorithm would still be the wrong tool for a one-year catastrophe and a Babel scattering. But it would no longer be allowed to spend 117,000 years inventing a crisis that Scripture never recorded and that the Y chromosome does not require.

Mainstream dates survive by treating the mutation rate as if it were a substitution rate stretched across imagined epochs. Jeanson has been explicit that he is doing the opposite: using the rate we can count in the present and asking whether the living tree fits the years Moses wrote down. It does. FitCoal never asks that question. That is why FitCoal and Genesis disagree.

The Better Reading of the Genome

Living human DNA is not a million-year archive of near-extinction. It is the record of a kind God made in His image, judged in the days of Noah, preserved through the Ark, and scattered into nations. Mutations since then are real. They are mostly the wear of a fallen world, the bondage to decay Paul describes in Romans 8:20–22, not the raw material of new creatures. Adaptation after the Flood used information already present in those eight survivors. It did not wait 117,000 years for a bottleneck to invent humanity.

The FitCoal paper is useful in one limited sense. It shows, once again, that if you hand a sophisticated model the evolutionary timescale, the model will hand it back to you dressed as a discovery. The assumptions are the theory. When those assumptions are removed, the 900,000-year bottleneck has nothing left to stand on.

Keep comparing the clocks. Keep reading Genesis as history. The same Lord who numbered the years from Adam to Noah has not left Himself without a witness in the families of the earth. The genome, read with a rate we can measure and a timeline He has spoken, still points home.