The short version

Somebody answers a claim about what generally happens with an account of what happened once. The account may be entirely true, told in good faith, and important to the person telling it.

It’s still a sample of one, and it was selected by the fact that it was memorable enough to be worth mentioning.

Why it doesn’t work

A single case establishes that something is possible. It cannot establish how often — and how often is nearly always the actual question. “Does this work?” almost never means “has this ever once worked.”

The size isn’t really the problem, though. The selection is. Anecdotes reach you because somebody found them worth repeating, and what makes a thing worth repeating is that it was surprising. The stories that travel are the ones that were unusual enough to travel, which means the mechanism delivering them to you is a filter for unrepresentativeness. The typical case is, by construction, the one nobody tells you about.

Then there’s how the two kinds of evidence land. A rate is a fact about a population — abstract, external, nothing to picture. A story is an experience you briefly had. Nobody is calibrated for that difference, and the story wins in a room every time, including in the head of the person who knows better.

And it functions beautifully as a conversation-ender, because it can’t be checked. You cannot sample more of it, you cannot look it up, and disputing it means implying somebody is lying about their own life — which they almost certainly aren’t. “It worked for me” is the strongest form there is: sincere, unverifiable, and entirely silent on the question actually being asked.

And there’s a third option that leaving out is what makes the whole thing feel unanswerable. An account can be sincere and still be wrong. Memory isn’t a recording played back; it’s reconstructed fresh each time, steadier on the gist than on what happened in what order and how long the gap was. But the bigger problem isn’t memory at all. “I took it and I got better” can be perfectly accurate as a report and still be silent on cause, because the complaints people reach for a remedy for are very often the ones that ease on their own. Colds end. Flare-ups subside. A bad stretch is usually followed by an average one. Whatever you did last before improving collects the credit, and there is no version of that week running alongside where you didn’t do it. Nobody is misremembering. They’re reading a cause off a sequence, and a sequence is the one thing that can’t supply one.

This is not the fallacy

• Existence claims. One case is a complete and sufficient proof that a thing can happen. “That’s impossible” dies to a single example, and rightly.
• Illustrating a statistic. A story attached to data makes the data legible, and people remember it. That’s communication doing its job, not evidence being substituted.
• When n=1 is the whole literature. Rare conditions, novel incidents, your own body. Sometimes the anecdote is genuinely all there is, and using it is correct — it’s the overclaiming from it that isn’t.
• A story offered as a question. “This happened to me — is that common?” is exactly the right thing to do with an anecdote.

The test is one line: is the claim about whether it can happen, or about how often? An anecdote settles the first completely and the second not at all.

How to call it out

This one needs more care than the rest of them, because you’re standing very close to somebody’s personal experience, and the fastest way to lose the argument is to appear to be calling them a liar:

• “I believe you completely. Do we know how typical that is?”
• “That’s one. Is there a rate anywhere?”
• “That’s a real thing that happened — I want to know how often it happens.”

What not to say is “that’s just an anecdote.” It’s dismissive, it reads as an accusation, and it’s also slightly wrong: their experience isn’t worthless, it’s just answering a different question from the one on the table. Grant it in full, and then ask for the number. Almost nobody will refuse.