How Dating Methods Can Fail

And Why They Fail Too Old

Hualalai, on the island of Hawaii, erupted in 1800 and 1801. We know this because people watched it happen and wrote it down.

In 1969 the geochronologist G. Brent Dalrymple measured the potassium-argon age of that lava. The rock came back at roughly 1.4 million years old.

This is not a fringe result or a contested one. Dalrymple published it himself, and he was one of the people who built the method. He was not embarrassed by it, because he understood exactly why it happened. The same study reported historically dated flows from Mount Etna at 140,000 to 350,000 years. Sunset Crater in Arizona, which erupted around 1065 AD, came back at about 270,000. The 1959 Kilauea Iki flow has returned ages between 1.7 and 8.5 million years. The 1915 eruption of Lassen Peak dated to about 110,000.

Every one of those rocks has a known birthday. Every one of them dated old — some by three orders of magnitude, some by five.

That is the fact this piece starts from. Not a theory about it. Just the fact.

What a Date Is Made Of

Nothing above is a failure of physics. Radioactive decay is among the most precisely characterized processes in science, and the decay constants for potassium-40, uranium-238 and carbon-14 are not in question here or anywhere else. The atoms decay at the rate they decay.

A radiometric age is two things joined together. The first is a measurement: how much of a parent isotope is present, how much of the daughter product. That measurement is real, reproducible, and not in dispute.

The second is an assumption about starting conditions. Potassium-argon dating works by counting the argon-40 that has accumulated since the rock formed — which requires assuming that the rock contained essentially no argon-40 to begin with. For a lava flow that erupted into open air and degassed as it cooled, that is often a fair assumption.

It is not a fair assumption everywhere. Argon dissolved in magma at depth can be trapped rather than released. Rock that cools too quickly has no time to degas. Water pressure holds gas in. Where any of that happens, the rock starts its life with argon already inside it, and the method counts that inherited argon as though it were the product of elapsed time.

The measurement was correct. The arithmetic was correct. The answer was wrong, because the premise handed to the arithmetic did not describe the rock.

Why the Error Runs One Way

Here is the part that matters more than any individual number.

Excess argon can only make a rock look older. There is no mechanism by which trapped argon makes a sample appear younger than it is — the extra atoms add to the count, and the count is what the age is calculated from. The error is not scatter around a correct value. It has a direction, and the direction is always the same.

Dalrymple and Moore found the sharpest case in 1968, dating fresh pillow basalt from the submarine flanks of Kilauea and Mauna Loa — rock of essentially zero age. The excess argon rose with water depth. It was highest of all in the glassy rims, where molten rock had quenched instantly against seawater and had no opportunity to degas at all. Fast cooling under water is the condition that traps argon most effectively.

That is worth holding onto, because it describes how new ocean floor forms.

The Dog That Was Born in 1899

The same structure appears in a place with no isotopes in it whatsoever.

The German Shepherd was founded as a breed in 1899. The date is documented; there is no inference involved. Compare German Shepherd genomes to wolf genomes using standard molecular-clock methodology, and the calculated divergence comes out too old by a factor of between 336 and 2,297, depending on which published mutation rate is used.

The mutation rate is not wrong. What is wrong is an assumption buried in the method: that the genetic differences between two populations accumulated as new mutations after they separated. In the dog case, 99.95 percent of the observed differences are ancestral variants — variation that already existed in the wolf population and was sorted into the breed by selective breeding. The clock counts every difference as a tick. Nearly all of them were there before the clock started.

And again the error has a direction. Sorting pre-existing variation can only inflate an apparent divergence. It cannot deflate one.

Two methods, unrelated physics, no shared machinery. Both fail against known ages. Both fail too old.

Disturbed, or Not

One question organizes everything that follows: was this material disturbed, or was it not?

That question does the sorting, and it is the reason this is a framework rather than a complaint. A position that simply distrusts dating explains nothing and risks nothing. A position worth testing has to say in advance which results it expects to be wrong and which it expects to be right — and then be held to both lists.

The event in question is set out in full elsewhere, and only its outline matters here. The rigid outer shell of the Earth fails along weaknesses it already had. The continents separate rapidly. Molten rock rises to floor the new ocean basins and quenches against flooding seawater — the same conditions Dalrymple and Moore found trapping argon in the Hawaiian pillow rims, across millions of square kilometers instead of one submarine flank.

The claim that follows is narrow and specific. For the material that event touched, the conditions the calibrations assume did not hold. Mantle degassing raised ambient argon. Molten rock quenched against flooding seawater. Carbon that had never seen the atmosphere entered the carbon cycle. Radiation backgrounds shifted as fresh volcanic material was spread across the surface.

For material the event did not touch, none of that applies, and the methods return what they have always returned.

What Still Reads True

This is the half of the argument that gets left out when frameworks like this one are summarized by their opponents, so it goes here rather than in a footnote.

Uranium-lead and isochron dating are not contested. These methods are specifically built to avoid the vulnerability described above — isochron techniques solve for the initial daughter ratio instead of assuming it, and uranium-lead uses two independent decay chains on zircon, a mineral that rejects lead when it crystallizes. Applied to old, undisturbed crystalline rock, they return that rock's true age. The framework expects them to.

Old rock is expected to read old. The floor of the Pacific is remnant sea floor that predates the event, and every robust method applied to it should return a large number. That is a prediction, not a concession.

Tree rings, cave formations, coral and lake varves read true. A tree-ring sequence spanning several thousand years is largely or entirely pre-event material, counted correctly. These records are not among the open problems.

The geological column stands. Superposition, cross-cutting relationships and faunal succession were all established before radiometric dating existed, and none of them depends on absolute numbers. Whatever the true ages turn out to be, everything stratigraphically lower remains lower and everything above remains above. The order is not at stake. Only the numbers attached to it are.

Old fragments turn up in young deposits, and that is expected. Catastrophic transport carries pre-existing rock with it. Dating a fragment returns the fragment's age, not the age of the bed it now sits in. A scatter of old ages on reworked material is the signature of transport, not a contradiction.

There is even a clean case where potassium-argon works on material of known age: sanidine crystals from the 79 AD eruption of Vesuvius return the correct calendar date. Sanidine crystallizes hot and degasses thoroughly. The method is not broken. It is conditional, and the conditions are nameable.

Where the Date Comes From

If the conventional numbers are artifacts for the disturbed material, some other measurement has to supply an actual date. One does, and it is worth being precise about how much weight it carries.

Every human genome contains a load of private variants — mutations present in one individual and effectively absent from everyone else. These are recent. They arose in the germline of that person's recent ancestors and have not spread through the population. Whole-genome sequencing puts that load at roughly 14,000 to 16,000 per individual. Separately, sequencing parents alongside their children measures the rate at which new mutations appear: approximately 70 per generation.

Divide one by the other. The result is 175 to 267 generations. At about 27 years per generation, that places the starting point between 4,725 and 7,200 years ago, centred near 5,800.

No effective population size enters that division. No parameter from any model in the series enters it. It is one measured quantity divided by another.

It is also one line of evidence, and the paper says so plainly. The full work sets out nineteen lines across seven disciplines that are consistent with that window — but only that one measures it. The rest are constrained by it: they are consequences worked forward from the date and from the physical model, and consequences agreeing with their premise is coherence, not confirmation. A framework that counted all nineteen as independent votes would be overselling itself.

But being constrained is not the same as being decorative. A theory whose single date forces nineteen consequences has nineteen places it can be killed, and it cannot repair any one of them without breaking the others. Independent results can each be wrong quietly. A chain cannot — an error anywhere surfaces somewhere else as a contradiction. That rigidity is what makes the thing testable at all, and it is why the tests at the end of the full argument are worth stating in advance.

What This Does Not Settle

Four full-scale computational models would carry the assessment past where reduced-order work can reach, and none of them has been run: a geodynamo simulation, a coupled climate model, an ice-sheet model, and a three-dimensional tectonic model. Each is specified in enough detail for someone with the tools to pick it up, and each could break the framework.

The depositional work was scored against real sites — 23 of 24 predictions on the Colorado Plateau, 4 of 5 on a blind test in South Australia chosen outside the authors' hands. Twenty-seven of twenty-nine is a score, not a validation, and the two misses stand.

One prediction was written down before it could be checked and still has not been checked: that per-individual private variant load should be roughly equal across human populations. The conventional model predicts otherwise. The data exist. The test has not been run.

The full correction curve for each dating method — the shape, not just the direction, of the error — has not been calculated. Only the direction is claimed.

The ice cores are the hardest case in the set, and they are not dodged. On this account the polar ice is entirely a product of the event: there was no perennial ice before it, which means the whole record has to fit inside the window rather than a recent cap of it. What the work supplies is a supply budget — whether enough moisture existed, and at what efficiency it would have to be delivered. Whether that supply becomes the observed ice, at the observed rate and no more, is one of the four models nobody has run.

The Rhinoceros in the Drawer

In October 2025 a rhinoceros was described from Devon Island in the Canadian Arctic, at 75° North: a nearly complete skeleton, recovered from lake sediments inside Haughton Crater, alongside a temperate-forest fauna at a latitude that today supports ice and rock. Its closest relatives are European. Three months earlier a separate team reported partial sequences from seven proteins recovered from its tooth enamel — a specimen roughly ten times older than anything from which endogenous DNA has been recovered, at a conventional age of 21 to 24 million years.

Its describers propose that a North Atlantic land bridge stayed crossable for twenty million years longer than previously thought, in order to get it there.

That animal is what started this project, and it is where the argument can be settled rather than debated. The preserved proteins contain organic carbon — carbon that was part of a living animal, carrying whatever ratio of carbon-14 the atmosphere held when the enamel formed.

At 23 million years, carbon-14 is gone. Not scarce: gone. The conventional framework predicts zero.

At a few thousand years, roughly half of it remains, and modern instruments detect far less than that.

The specimen is in a museum collection. The organic carbon has already been confirmed present. One measurement, on one tooth, distinguishes between the two answers — and it is a measurement that can come back against this framework as easily as for it.

The rhinoceros was never the proof. It was the question.

This piece is the short way in. The full argument — nineteen lines of evidence, a method-by-method treatment of the conventional chronology, and twenty-one tests specified in advance — is set out across the papers below.

The Dating Capstone — the whole case, in four parts, with the tests.

What Broke the Foundations? — the physical mechanism. What kind of event this would have to have been. (Or start with The Cracked Shell, the shorter door.)

How Did Humanity Diversify? — where the date comes from, and the genetic work behind it.

When Did the Wolves Start Howling? — the German Shepherd test, and what molecular clocks actually count.

How Did the Rhino Cross the Sea? — the first paper, and the fossil that raised the question.

© 2026 D. L. White. Licensed under CC BY-ND 4.0. https://creativecommons.org/licenses/by-nd/4.0/

AI Collaboration Disclosure: Drafting and calculations by Claude (Anthropic), with adversarial review by Grok (xAI), under the direction of D. L. White. Neither AI system endorses all conclusions as settled.