Poisoning the well

“Take it with a grain of salt”

Before a team meeting, the team lead says: “Riley’s going to pitch the new scheduling software next. Riley gets excited about every shiny tool, so take whatever you hear with a grain of salt.” Riley then presents the price, a trial run’s results and three other teams’ experiences.

Everything Riley presents can be checked: the price is the price, and the trial either saved time or didn’t. The warning doesn’t touch any of it. What it does is tell the room to discount all of it, in advance and without looking. If Riley’s enthusiasm has led to bad calls before, the fair response is to scrutinize the numbers, not to discount them unseen.

A well poisoned for anyone who disagrees

A board game club’s longtime champion explains his favorite opening strategy to new members, then adds: “Anyone who tells you it doesn’t work is just a sore loser who couldn’t pull it off.”

This version targets a whole group, in advance, before any of them has spoken. It’s also self-sealing: anyone who later reports that the strategy failed for them becomes, by definition, a sore loser, and their report counts as confirmation. Whether the opening works is a question about results, and it has been set up so that no result can count against it.

A true fact used as a blanket dismissal

Sam writes a review of the office’s cleaning contractor, listing six missed visits with dates and photos. Before circulating it, a manager adds a note: “Worth knowing that Sam had a falling-out with their supervisor last year, so I wouldn’t put much stock in this.”

The falling-out may be true, and it’s a fair reason to check that Sam isn’t exaggerating. But the dates and photos can be checked against the contractor’s own visit log. The note takes a real reason for scrutiny and uses it as a reason to skip scrutiny. That’s what makes this form persuasive: the fact it rests on is accurate, so the dismissal sounds like caution rather than prejudice.

• Blanket character attack: presenting someone as so dishonest, fanatical or dogmatic that nothing they say can be trusted. Erik Krabbe and Douglas Walton call this “an extreme form of ad hominem attack which appears to leave the attacked party no room available at all for defence.”
• Advance charge of bias: announcing before someone speaks that their interests or allegiances will decide what they say. Krabbe and Walton note that some writers use “poisoning the well” for every accusation of bias, but find it useful to set apart the harshest form, where the accusation is meant to rule out anything the person might say.
• Self-sealing: framing things so that a denial or a reply confirms the charge, as in the board game example. This structure is what makes the charge unfalsifiable. The Kafka trap, in which denying an accusation is taken as evidence for it, is a close relative.

Where the name comes from. The phrase is John Henry Newman’s. In 1864 the novelist Charles Kingsley, in a public dispute over whether Newman was truthful, wrote that he was “henceforth in doubt and fear, as much as any honest man can be, concerning every word Dr. Newman may write.” Newman replied in his Apologia pro Vita Sua that Kingsley had tried “to poison by anticipation the public mind against me... and to infuse into the imaginations of my readers, suspicion and mistrust of everything that I may say in reply to him. This I call poisoning the wells.” Newman’s point was that if the charge took hold, answering it would be a waste of time, “and this is precisely what he knows and intends to be its fruit.” Logic textbooks later adopted the case: Walton notes that Irving Copi and Carl Cohen’s textbook uses it to illustrate the fallacy.

Its place in the family. Poisoning the well is classified as a type of ad hominem. Krabbe and Walton treat it as “a particularly harsh subspecies” of the bias type, and Walton’s book on ad hominem arguments gives it a section of its own as a subtype. Walton also notes that textbooks differ on whether it belongs under the bias type or the circumstantial type. Krabbe and Walton consider other types of ad hominem, such as direct, circumstantial and bias attacks, reasonable in some contexts, but they judge poisoning the well always fallacious in a discussion aimed at resolving a disagreement, because it ends the discussion instead of advancing it. It differs from tu quoque in timing: a tu quoque answers what someone has already said, while poisoning the well gets in before they say it.

Krabbe and Walton argue that an accusation of bias can be a reasonable move in a discussion. It raises a real question and puts the other person under pressure to answer it. What turns it into poisoning the well is using it to close the discussion instead. Advance information about a source is legitimate when:

• It’s specific and bears on reliability for this question: a financial stake in the answer, a track record of errors on this subject, a documented reason to doubt this kind of claim.
• It’s offered as a reason to check, not a reason to stop listening. The claims still get heard and weighed.
• It leaves room to respond. Krabbe and Walton stress that the other person must still be able to answer, for instance by pointing to independent evidence.
• You’ll be relying on the source’s word, because the claims can’t be checked any other way. Then knowing about the source in advance is useful.

The test: does the warning tell you what to check, or tell you not to bother listening?

The editor’s note

A cycling magazine publishes a review of a new bike light. Above it, the editor notes: “The manufacturer provided this light to our reviewer free of charge. The reviewer’s test data, including battery life and brightness readings, are in the table at the end.”

This tells readers something about the reviewer before they read a word of the review, which is what poisoning the well does. But the fact disclosed is specific and relevant, it doesn’t tell anyone to disregard the review, and the note directs readers to the data they can check. Disclosure is also what separates this from the damaging kind of bias charge: Krabbe and Walton’s example of a powerful attack on someone’s sincerity assumes she had not told the others about her interest at the start. This meets the reason to check and room to respond conditions.

The used car

Before a friend goes to see a used car, you tell her: “The seller’s ad says it’s never been in an accident, but the vehicle history report shows two insurance claims for body damage. Whatever he tells you, get it inspected.”

You’re discrediting a source before your friend has heard from him. But the warning rests on documented evidence about the very claim he’s making, it’s aimed at a claim she can’t easily check herself, and it tells her how to find out rather than telling her what to conclude. If the inspection shows no damage, nothing you said stops her from buying the car.

First impressions shape how everything after them is read. Newman described this in his reply to Kingsley: “We interpret it by our antecedent impressions.” A defense, he wrote, will be read “in a good sense or a bad” depending on what the reader already believes about the person making it. Once listeners expect dishonesty, ambiguous statements look evasive and ordinary caution looks like spin, in the same way that a bad overall impression of a person spills over into other judgments of them (see the Halo effect). Newman put the result sharply: “The very same sentiments, according as our jealousy is or is not awake, or our aversion stimulated, are tokens of truth or of dissimulation and pretence.”

It’s also efficient. Answering claims one at a time takes work; discrediting their source once disposes of all of them, including ones not yet made. And because judging claims by their source is often sensible (see Genetic fallacy), an advance warning sounds like prudence.

• Ask what, specifically, the warning is based on, and whether it bears on the claims about to be made.
• Hear the claims anyway, and check the ones that can be checked. A reason for suspicion is a reason to look harder, not to look away.
• Watch for a charge that no answer could satisfy. If denial counts as proof, the charge has stopped being evidence.
• If you’re the target, Krabbe and Walton suggest acknowledging any real interest and pointing to evidence that doesn’t depend on your word. Newman’s own response was to name the tactic plainly before answering the charges.