Testing a claim

The teaching

As the craft of the historian, the reporter and the fact checker teaches it.

Find the first telling

Every claim has a first teller, and finding them is the first test. Search for the earliest version you can, and note who said it, when, and what they pointed to. Claims change as they travel: a guess becomes a fact, a maybe becomes a certainty, a name is added, a date is moved. A quotation needs a book and a page, or a recording with a date; a quotation that no one can place, however often it is repeated, is not yet evidence of anything.

Read the source whole

When a claim rests on a document, a film or a figure, go to it and read or watch the whole, not a clip or a screenshot. A sentence cut from its paragraph can mean its opposite; a chart can start its axis wherever it likes; a photograph can be real and its caption false, which a reverse image search often shows by finding the picture years earlier in another story. A translation should be checked against the original or a second translator. If the source cannot be found at all, that is itself a finding.

Read across, not down

Professional fact checkers do not judge a website by reading it closely, as students are taught to. They leave it almost at once, open new tabs and find out what others say about the source: who runs it, who funds it, what its record is. A study of 2017 found that this habit, called lateral reading, let fact checkers reach sound judgements faster than historians or able students who read the page itself. One simple method built on it, called SIFT, has four steps: stop before believing or sharing; investigate the source; find better coverage of the claim; and trace claims, quotations and images back to their original context.

Say what would prove it wrong

Before looking at the evidence, write down what result would change your mind. For a claim about the world, something must be able to count against it, or it says nothing: a theory that fits every possible finding explains none of them. If, when a finding comes in against the claim, the claim is quickly changed to fit, the goalposts have moved, and the claim was never being tested.

Count the keepers of the secret

List everyone who would have to know and keep quiet for the claim to be true, and for how long: officials, technicians, contractors, their families, their successors and their rivals. Real conspiracies are kept by small circles for limited times, and even they leak, through a conscience, a quarrel, a divorce, a deathbed, a lost file or a lawsuit. A secret that would need hundreds of thousands of people, across rival countries, for generations, without one credible defector or document, asks for more than human beings have ever managed.

Weigh the plainer story

Of two explanations that fit the facts equally well, prefer the one that needs fewer assumptions. This is Occam's razor, a rule of thumb and not a proof, but it sets the order of inquiry. Mistakes, muddle and ordinary self-interest explain much that looks planned. And coincidence is commoner than it feels: among eight billion people, events with odds of a million to one happen thousands of times a day, so a strange coincidence alone proves nothing.

Who benefits proves nothing

Conspiracy theories often begin from the question who benefits, cui bono. It is a fair question for an investigator choosing where to look, but it is not evidence. Many people profit from events they did nothing to cause: the undertaker from a death, the builder from a storm, a rival from a scandal. Benefit gives a motive to examine, and then the work of finding what was actually done begins.

Proportion belief to the evidence

“A wise man, therefore,” wrote David Hume in 1748, “proportions his belief to the evidence.” Belief is not a switch but a dial: a claim can be likely, possible, doubtful or nearly ruled out, and it moves as evidence comes in. Laplace added that the stranger a claim, the stronger the evidence it needs, since it must outweigh everything else we know. Holding a claim as unproven is not the same as calling it false; it is an honest place to wait.

The marks of a claim that cannot lose

Some ways of arguing show a claim protecting itself rather than being tested. Questions are asked but never answered by looking, only multiplied. Claims come in a flood, too many to check, so that an answer to any one is waved away for the rest. Evidence against the claim is taken as proof of a cover-up, and experts are called bought unless they agree. And the call to do your own research turns out to mean reading only the theory's own sources. Each lets a claim survive without ever risking being wrong.

One claim tested: the Moon landings

Take the claim that the Apollo Moon landings of 1969 to 1972 were filmed on Earth. Its first full telling is a self-published book of 1976 by Bill Kaysing, who had written technical manuals for a rocket company, and it spread through a television special in 2001 and then online. It points to the photographs: a flag that seems to wave, a black sky with no stars, shadows that are not parallel.

Each has a plain answer. The flag hung from a horizontal rod and moved only while it was being handled, swinging longer because there was no air to still it. The sky shows no stars because the cameras were set for a brightly lit surface, as a photograph of a floodlit stage at night loses the stars. Shadows on uneven ground, cast by a single sun, fall at different angles.

Then count the keepers of the secret. About four hundred thousand people worked on Apollo. The Soviet Union, racing the Americans and tracking the missions, had every reason to expose a fake and never claimed one. The three hundred and eighty-two kilograms of rock brought back have been studied by laboratories around the world and show features no rock formed on Earth has. Reflectors left by Apollo 11, 14 and 15 still return laser pulses fired from observatories, and an orbiting camera photographed the landing sites, with their descent stages, tracks and footpaths, in 2009 and 2011. The claim fails every test above, and the record is open for anyone to check.