Checked · The Younger Dryas

What would actually settle the Younger Dryas argument

The claim

The Younger Dryas impact argument cannot be resolved, so it comes down to which side you trust.

Our verdict

Refuted

It is resolvable, and both sides could agree on the test in advance. Pre-register a marker, a set of sites, a measurement protocol and a threshold, then have laboratories that have never published on this run it blind. The reason that has not happened is not that the question is unanswerable.

Eighteen years, dozens of papers, and four rounds in one journal between 2021 and 2024. It is reasonable to conclude the argument cannot be settled. That conclusion is wrong, and saying why is the most useful thing this page can do.

Why the current exchange cannot resolve it

Every round works the same way. One side reports markers at a set of sites. The other analyses samples and does not find them, or finds them and attributes them differently. Then each side explains why the other's method was wrong.

That structure cannot converge, because nothing was agreed in advance. What counts as the boundary layer, which sites are diagnostic, which extraction protocol is valid, and what concentration counts as anomalous are all decided after the samples are in, by people who already know what they expect to find.

This is not an accusation. It is the ordinary condition of a field argument, and it is why medicine invented pre-registration.

The test

Four things fixed in writing, before anybody opens a sample.

  • One marker. Not a suite. Platinum is the strongest candidate, because it is measurable to low concentrations by standard methods and both sides have published on it.
  • A site list, agreed by both sides. Including sites each side expects to be negative. A protocol that only samples where the signal is claimed cannot fail.
  • One extraction and measurement protocol, written out, with the depth sampling and the calibration specified.
  • A threshold and a decision rule. What concentration, over what background, at what fraction of sites, counts as a positive. Written down before the numbers exist.

Then send blinded, coded samples to laboratories that have never published on the Younger Dryas, and publish the result whichever way it falls.

Candidate

This is doable now with existing equipment and no new science. It would cost a fraction of what has already been spent arguing, and it would produce a number that neither side could dismiss as somebody else's bad method, because both would have signed the method.

Our proposal. It has not been run, and to our knowledge nothing of this design has been proposed for this question by either side.

What each side has to give up

What the proponents accept

A negative result at agreed sites, on an agreed protocol, counts against the hypothesis, and cannot be answered by proposing that the marker was not the right marker.

The site list includes places where the signal is not claimed.

The threshold is fixed before the data exists.

What the sceptics accept

A positive result at agreed sites counts for the hypothesis, and cannot be answered by proposing contamination after the fact.

The protocol is the one they signed, so a failure to reproduce is not available as a method critique.

The question is treated as open enough to be worth testing, which some of the refutation language does not concede.

Both columns are uncomfortable, and that discomfort is the point. A test both sides find comfortable is a test that cannot change anybody's mind.

The same request, in three fields

This is the third time on this site we have asked for the same thing. A false-positive rate for the Indus script entropy measure. A character error rate for the Herculaneum ink detection. A pre-registered threshold for the Younger Dryas markers.

Three unrelated disciplines, and the missing number is the same shape in all three: the one that says how often the method is wrong. We ask it of ourselves too, and we publish the answer. Our lidar detector recovers 33.1% of scheduled monuments on Salisbury Plain and 28.7% on Bodmin Moor, and its precision on Bodmin is 5.7%.

Where this leaves things

The five labels, applied to this question. Nothing is left unlabelled.

Refuted

The claim that this argument is unresolvable in principle. It is resolvable by a test that requires no new science.

Candidate

The pre-registered platinum protocol proposed here. Ours, untested, and offered so it can be improved or dismissed on its merits.

Under review

The hypothesis itself, pending something of this kind.

Published

The four-paper exchange, which is real, readable and worth following in full.

Speculative

Any prediction about which way a blind test would fall. We have none.

Sources

Related

Where this is argued in full

Answering the Architect Question

Both cases at the same length, measured and printed at the back. The argument runs across four papers in one journal and you can read all of it.

Both cases at the same length, measured and printed at the back. It is written and not on sale, and advance readers can read it first.