HistoricalChristian.Faith

A recent eight-person bottleneck predicts a detectable genetic signature — is it there?

Type: supporting argument (off-timeline) · Assessment: weak — as a prediction of the literal-eight model dated to the last several thousand years. This is the sharpest test on the timeline, because the model's genetics are unusually precise: eight passengers, of whom only five contribute distinct ancestry, cap the surviving human gene pool at ten alleles per autosomal locus, four maternal lineages and one paternal one. That is a number, and numbers can be checked. The check fails at the date rather than at the idea: the coalescence times of human genetic variation are orders of magnitude older than a few thousand years, and no crash to a founding handful appears anywhere in the recent record. Two honest corrections cut the other way and are stated here rather than buried — the much-repeated "human population never dropped below 10,000" claim was overstated and has been walked back by the people who made it, and the one dramatic bottleneck genetics has recently reported points away from the model in both directions at once, being far older and far larger than it predicts, and is itself disputed. The failure is scoped to the young literal-eight model, not to a flood tradition with a historical kernel.

Related arguments: testing the premise forward — see also a synchronous worldwide sediment horizon · single-point post-Flood faunal radiation · convergent independent flood narratives · authenticated Nephilim material remains · independent traditions of a long-lived first age

The prediction

Genesis 7:13 names the passengers: Noah, his wife, his three sons, and their three wives — eight people, and the text is emphatic that they were all. Take that literally, and place it a few thousand years ago, and the genetic consequences are not vague. The three sons carry only their parents' alleles, so the founding pool is five genetically distinct individuals: Noah, his wife, and three daughters-in-law. That caps the whole species at ten alleles at any autosomal locus, four mitochondrial lineages (the four women), and one Y-chromosome lineage (Noah). Everything human beings now carry must have arisen from that ceiling by mutation in the intervening millennia.

The prediction follows: modern human genomes should show a sharp, datable coalescence crash — an abrupt convergence of nearly all lineages onto a handful of ancestral haplotypes at a horizon a few thousand years back, with a correspondingly abrupt collapse of linkage disequilibrium and allele counts across the genome. It is a good prediction. It is risky, it is quantitative, and it is testable with methods designed for exactly this question.

What would falsify it — committed in advance

The prediction fails if the times to most recent common ancestry across many independent loci come out far older than the proposed horizon, and if no genome-wide signature of a severe recent contraction can be recovered. It fails in particular if some locus can be shown to carry more distinct lineages, of greater age, than eight founders could have transmitted — since that is an arithmetic contradiction rather than a probabilistic one.

It would be confirmed by a genome-wide coalescence floor at the right date. And it is worth saying plainly that this is not an unfalsifiable demand: severe bottlenecks are detectable and have been detected. The out-of-Africa founder effect is written across non-African genomes as reduced diversity and long haplotype blocks; the founder events in Ashkenazi, Finnish and Icelandic populations are recoverable in detail. The instrument works. The question is only what it registers at the proposed date.

The strongest case for

The mitochondrial and Y-chromosome data have the right shape, if not the right scale. All living human mitochondrial lineages do coalesce to one woman, and all Y lineages to one man — the "Mitochondrial Eve" and "Y-chromosome Adam" of the literature. That is a real and striking result and it is what the model expects qualitatively. Nathaniel Jeanson has pressed the quantitative form of it, arguing from directly measured (pedigree) mitochondrial mutation rates rather than fossil-calibrated ones, and getting a mitochondrial coalescence far younger than the conventional figure.¹

And the standard rebuttal has been overstated, which the argument is entitled to point out. For years the reply of choice was that population genetics shows human numbers never fell below about ten thousand, so a founding couple or a founding eight is excluded outright. Richard Buggs, an evolutionary biologist, challenged the inference on method: the estimators involved measure long-term effective population size and assume relative stability over many generations, and are therefore poorly suited to detecting a short, sharp crash followed by rapid recovery. The exchange that followed was long and public, and it ended with the critic largely vindicated on the narrow point — Joshua Swamidass's analysis put the boundary at the "time to most recent four alleles" (TMR4A), a median of roughly 500,000 years, meaning a bottleneck of two is excluded by genetics only within that window and not before. BioLogos subsequently revised its position, and the articles making the stronger claim were withdrawn. The argument's complaint that it was met with more confidence than the data supported is justified.²

A second overreach on the same side is worth recording, because it went the same way. Francisco Ayala argued in 1995 from the extreme polymorphism of the immune-system HLA-DRB1 locus that its lineages coalesce some 60 million years ago, before the ape–monkey split, and that human ancestral populations therefore can never have been smaller than a few thousand. That argument was answered within genetics: Bergström, Josefsson, Erlich and Gyllensten reported in 1998 that the DRB1 alleles are of recent origin, undercutting the old trans-species coalescence dates the minimum-population inference rested on. A page that leaned on Ayala here would be leaning on a prop that has already broken.³

Adversarial assessment

What holds. The universal mitochondrial and Y coalescences are real. The methodological critique of the "never below 10,000" claim is correct and was conceded. The window in which a two-person bottleneck is genetically excluded is bounded, and the boundary is around half a million years, not the whole of human history. Anyone repeating the older, harder claim is repeating something that has been withdrawn.

But every one of those concessions is about a date the model does not want. This is the decisive point and it should not be softened. TMR4A of roughly 500,000 years is not a licence for a recent bottleneck; it is the opposite. It says that if a founding pair existed they must be placed before that horizon, because any more recent date would require them to have carried more alleles than two people can carry. The literal-eight model relaxes the constraint slightly — ten alleles rather than four, four maternal lineages rather than one — but it relaxes it by a factor of a few against a discrepancy of a factor of a hundred, and it places its founders not before 500,000 years ago but a few thousand. The concession the argument won runs directly against the model this page is testing.²

The mitochondrial and Y coalescences are also far too old, and they do not coincide. The mitochondrial figure runs to roughly 150,000–200,000 years and the Y figure to a comparable but distinct range; neither lands near a few thousand, and they are not contemporaneous, which a single founding family requires them to be. The pedigree-mutation-rate argument that would compress them is the pivot of the case and it is where the case is weakest: pedigree rates measure mutations transmitted over one or two generations, including transient heteroplasmies that do not fix, and applying them without correction across thousands of generations systematically underestimates ages — the discrepancy between pedigree and phylogenetic rates is a known problem in the field with a known direction, not an unnoticed inconsistency waiting to overturn the dating. And even if the mitochondrial clock were recalibrated, mitochondria are one non-recombining locus. The autosomal genome contains tens of thousands of independent histories, and it is those, not mtDNA, that carry the arithmetic ceiling.

The one dramatic bottleneck recently reported is the wrong bottleneck, and it is disputed. Honesty requires both halves. Wangjie Hu and colleagues inferred in Science in 2023 a severe ancestral crash to roughly 1,280 breeding individuals lasting some 117,000 years, at c. 930,000–813,000 years ago. That is a striking result and it is sometimes reached for as support. It is not support: it is about a thousand times older than the model's date and a hundred times larger than its founding group — a different event of a different kind. And it may not be an event at all. Yuxin Deng, Rasmus Nielsen and colleagues argue it is likely a statistical artifact, noting that no comparable bottleneck appears in non-African populations, and a second group reproduced the signal with the original method but not with an independent method on the same data. An argument that cites Hu is citing something contested, and contested in a direction that does not help it.⁴

And the one popular bottleneck of roughly the right magnitude did not survive testing either — a case worth recalling as calibration. Before the Hu result, the go-to "near-extinction" datum was the Toba super-eruption of about 74,000 years ago in Sumatra, the largest explosive eruption of the Quaternary, which Stanley Ambrose proposed in 1998 had thrown the world into a volcanic winter and crashed the human population to a few thousand — low enough, he argued, to drive rapid racial differentiation. For a while this was cited as the human bottleneck, and it is instructive here precisely because it failed on evidence the model's own defenders would grant is fair. Archaeologically, Michael Petraglia and colleagues recovered Middle Palaeolithic tool assemblages in continuity above and below the Toba ash at Jwalapuram in southern India — people living on both sides of the layer, not extinguished by it. Environmentally, subdecadal phytolith and charcoal records from Lake Malawi show no volcanic winter and no vegetation collapse in East Africa after the eruption. Genetically, no bottleneck coincident with 74 kya can be recovered at all; the genuine Late Pleistocene contraction is the later out-of-Africa founder effect, a different event of a different date. The Toba bottleneck is now widely rejected. Two lessons follow for this page: a proposed crash to a few thousand at 74,000 years is already more than ten times too old and far too shallow for a Holocene eight, and even that claim did not hold up — which is exactly what it looks like when a bottleneck signature is searched for and is not there.⁵

The undisputed part is the negative, and the negative is what the prediction asked about. Across all of this dispute — the ten-thousand claim, the HLA claim, the TMR4A boundary, the Hu bottleneck and its critics — nobody on any side reports a crash to a founding handful within the last several thousand years. That is not a gap in the search; it is the one thing the search agrees on.

A note on method. None of this assumes that God could not have acted. A miracle can be granted; what it cannot do is relocate a coalescence. If the objection is that God created the surviving eight already carrying, or subsequently given, the full diversity of the modern species, then the model has stopped predicting a genetic signature and is instead explaining why there is none — which is a coherent theological position and an abandoned scientific prediction. This page assesses the prediction that was made. Possibility is not proof, and a claim that survives every possible observation is not being tested by any of them.

What the failure does and does not touch. It falsifies the literal-eight-survivors model placed a few thousand years ago. It does not touch a regional flood with human survivors outside the affected area; it does not touch a genealogical reading on which a real couple are ancestors of everyone alive without being everyone's sole genetic source — a position argued at length by Swamidass and left standing by all of the above. And it says nothing about the antiquity of the flood tradition, which rests on the cuneiform record, not on genomes.

Assessment

Weak — as a prediction of the young literal-eight model. The model's genetics are admirably precise and therefore genuinely testable: five distinct founders, ten alleles per autosomal locus, four maternal lineages, one paternal, a few thousand years ago. What the genome shows instead is coalescence depths two orders of magnitude older, mitochondrial and Y ancestries that are both far too old and not contemporaneous with each other, and no recent crash to a founding handful anywhere in the data.

Two things must be conceded on the way, and are not concessions the argument can spend. The "human numbers never fell below ten thousand" reply was overstated, was challenged on sound methodological grounds by Buggs, and was withdrawn; Ayala's HLA argument for a minimum population was answered within genetics in 1998. But the boundary that replaced them is TMR4A at roughly half a million years, which excludes a small founding group precisely in the window the model needs and permits one only far outside it. And the one severe bottleneck genetics has lately reported is a thousand times too old, a hundred times too large, and disputed as an artifact besides. The prediction was well made and it did not come in. The failure belongs to the young literal-eight model, and to nothing else on this timeline.

One caveat is owed on the other side and is stated rather than buried: molecular-clock dating is the most assumption-laden method this reconstruction leans on, resting on a mutation rate, a generation time and a demographic model, with correspondingly wide intervals. That is weighed at are these witnesses securely dated? — where the openness is conceded and found not to help, since the disputed date runs three orders of magnitude in the wrong direction.

Sources