The work everybody skips

Reading the footnotes

Both sides of this argument talk about the citations without opening them. One side says the books cite real papers, which is true. The other says it does not matter, which is lazy. There are ninety-one distinct journal references across the two texts, and the only way to know what they are worth is to pull them one at a time and read what they actually say next to what they are said to say.

Scope, stated plainly

Six of ninety-one.

This is a sample, not an audit. Six were chosen for carrying weight rather than at random, which biases toward the load-bearing end and should be read that way. Eighty-five are still sitting there unread. What follows is the beginning of a job, and saying so is part of doing it honestly.

91Distinct journal references across both books

6Pulled and read so far

2Support the claim as used

2Point at something stronger than the use they were put to

The two that hold

Read them, and they do the job.

How fast a magnetic field can change

Nature Communications, 2020

Used for → the field can fully reverse within a human lifetime
Known The paper is Davies and Constable, "Rapid geomagnetic changes inferred from Earth observations and numerical simulations". It finds directional change reaching about ten degrees a year, roughly a hundred times faster than anything happening now, and it specifies that these bursts happen at times when the field strength is falling. The book, if anything, undersells it. The condition the paper attaches the fast changes to is the condition we are currently in. Checks out, and stronger than the use.

Cosmic rays and explosive eruptions

Gondwana Research, 2011

Used for → cosmic rays trigger explosive volcanic activity by affecting magma viscosity
Known Disputed Real, and the paper is bolder than the sentence citing it. "Explosive volcanic eruptions triggered by cosmic rays: volcano as a bubble chamber." The mechanism is not viscosity, though: it is muons from cosmic rays seeding bubbles in silica-rich magma that is already holding more water than it comfortably can. The paper then proposes that the 1991 Pinatubo eruption may have been set off by a typhoon lowering the air pressure and letting more cosmic rays through, and speculates that Snowball Earth followed volcanism driven by cosmic rays from nearby supernovae. The claim holds. The mechanism is described wrongly in one word, and the paper argues something considerably more dramatic than it was cited for.

The two that drift

Real papers, different claims.

The three thousand and six thousand year solar cycles

Astrophysical Journal, 2021

Used for → the longer superflare cycles on the Sun are 3,000 and 6,000 years
Unsupported The paper is Okamoto and colleagues, "Statistical Properties of Superflares on Solar-type Stars", using the full Kepler dataset. It is a good paper and it belongs in this argument: it is part of the body of work establishing that stars like ours flare far harder than the textbooks used to allow. What it does not do anywhere is establish a three thousand or six thousand year cycle on the Sun. That claim is carried by the sentence, not by the reference under it.

Fifteen-hundred-year cycles back to the Triassic

Scientific Reports, 2022

Used for → the 1,500-year cycles run back as far as the Triassic
Unsupported Two problems, both checkable in a minute. The paper finds the 1,500-year rhythm in the Jurassic, about 155 million years ago, not the Triassic. And the authors propose it comes from a monsoon-like system on Earth, not from the Sun, while the passage citing it is arguing that the Sun drives the cycle. The finding itself is genuinely striking, because this rhythm had never been found before the Quaternary. It just does not say what it is being used to say. Note also that this paper carries a published author correction, which is worth checking before leaning on any figure in it.

The two that are more interesting than their use

Cited for the wrong thing, and worth more than that.

These are the finds. In both cases the sentence in the book is not carried by the paper underneath it, and in both cases the paper contains something the argument could have used harder.

The Heinrich-Bond cycle

Quaternary Science Advances, 2021

Used for → the cycle now has a name, the Solar Heinrich-Bond cycle
Known Disputed The name is really there. Viaggi, on astronomical and sun-related cycles in the Antarctic record, proposes a Heinrich-Bond cycle of about 6,000 years. So the claim that it has a name in the literature is correct, and that is worth knowing because it is the sort of claim that gets waved away. What the citing passage leaves out is the size. The paper puts the sun-related cycles together at roughly three to seven per cent of the up-and-down the records show. A named minor contributor is a real thing and it is not a driver, and quoting the name without the number is where this one goes wrong.

Superflares and extinctions

Astrophysical Journal, 2017

Used for → the 1859 flare caused temperature spikes of three to five degrees
Unsupported The paper does not say that. It is Lingam and Loeb, "Risks for Life on Habitable Planets from Superflares of Their Host Stars", and it is about something else entirely. And what it is about is better. The authors argue that the most powerful superflares are plausible drivers of extinction events, and that the periodicity of those superflares corresponds to patterns in the terrestrial fossil diversity record. That is a mainstream astrophysics journal proposing that a solar cycle of outbursts shows up in the fossil record of life on Earth. It was sitting under a sentence about an 1859 temperature spike, which is a much smaller thing to have said.

What the sample suggests

Four ways a citation goes wrong.

Nothing in this sample is invented. Every reference is a real paper in a real journal, and several are good ones. The failures are all of a subtler kind, and they are worth naming because they are what an opponent will find.

The claim outruns the source. A real paper on superflares placed under a sentence about a cycle it never mentions. The reference lends credibility it did not earn.
The detail slips. Jurassic becomes Triassic. A monsoon becomes the Sun. Bubble nucleation becomes viscosity. Small errors, each one checkable in about a minute, each one enough to lose an argument.
The magnitude goes missing. The name of the cycle is quoted and the three to seven per cent is not. This is the most damaging kind, because everything stated is true.
The source is stronger than the use. Twice in six. A paper linking superflare periodicity to the fossil record, and a paper proposing supernova cosmic rays triggered a global glaciation, both cited for smaller things.

Two of six hold as used. Two say something different. And two are worth more than they were spent on.

Which is the useful finding

The common accusation is that the references are decoration. On this sample that is wrong: they are real, they are relevant, and the field they point to is livelier than the dismissal assumes. The actual problem is looser and harder to defend against, which is that the sentence and the source drift apart in small ways that anyone can check. Every one of the errors found here would take a determined opponent under two minutes to surface, and any one of them would be used to discard the rest.

And what would finish the job

Eighty-five references still unopened. It is slow work and it needs no equipment: pull the reference, read what it claims, read the sentence it was attached to, and write down which of the four things above happened. Done properly it would produce the single most useful document in this whole argument, and it would be useful to whichever side turns out to be right.

The wider source ledger →