The more detail, the less likely.

In 1983 the psychologists Amos Tversky and Daniel Kahneman gave people the description of Linda above and asked them to rank how probable various statements about her were. Buried in the list were the two that mattered: “Linda is a bank teller”, and “Linda is a bank teller and is active in the feminist movement”. More than eight in ten people said the second was more probable than the first — and it made no difference whether they were undergraduates or doctoral students in statistics.

It cannot be. The feminist bank tellers are a subset of the bank tellers: every last one of them is a bank teller, but plenty of bank tellers aren’t feminists. So “bank teller” sweeps up everyone that “bank teller and feminist” does, and more besides. In the language of probability, P(A and B) can never exceed P(A) — adding a condition can only hold the probability the same or push it down. Ranking the combination higher isn’t a close call or a matter of interpretation; it breaks the most basic rule there is.

The two statements aren’t rivals. One contains the other. The only question the slider asks is how big the contained slice is — and the answer is never “bigger than the thing containing it”.

Faced with a hard question — how probable is this? — the mind quietly swaps in an easy one: how much does this resemble the story? Kahneman and Tversky called the shortcut the representativeness heuristic. Linda’s description was built, detail by detail, to sound like a feminist and nothing like a bank clerk. So “feminist bank teller” feels like a snug fit and “bank teller” feels wrong — and the feeling of fit gets reported as a judgement of probability.

That is why the extra detail backfires so reliably. Each specific you bolt on — and a feminist, in Portland, who owns a cat — makes the picture more lifelike and easier to believe, while quietly multiplying the conditions that must all hold at once. Vividness climbs; probability falls. The story and the odds move in opposite directions, and our intuition tracks the story.

This is not just a party trick with an invented woman. Forecasts get more believable and less likely the more detail they carry: Tversky and Kahneman found people judged “a Russian invasion of Poland, and a suspension of diplomatic relations with the USA” more probable than the plain “a suspension of diplomatic relations” — because the invasion supplies a satisfying reason. A concrete, well-told scenario always beats a vague one for persuasiveness, and always loses on probability.

Insurance runs on it: people will pay more for a policy covering “death due to terrorism” than one covering “death from any cause” — though the second includes the first. Juries and diagnoses feel the same pull, finding a detailed, coherent account of events more credible than the bare, broader one that must actually be at least as likely. Wherever a rich narrative competes with a plain possibility, the conjunction fallacy is quietly charging a tax on the good story.

“It’s more likely because it makes sense” is the whole error in one sentence. Making sense is about fit; likelihood is about counting.

Look for the nesting. Ask whether one option is the other with an extra clause attached. If it is, it is the smaller set and cannot be more probable, however much better it sounds.

Strip the detail and re-ask. Delete the vivid specifics and pose the plain question: how likely is the broad thing? The narrower version is a fraction of that answer, never more.

Count in frequencies. Reframe “how probable” as “how many out of 100”: of 100 people like Linda, how many are bank tellers, and how many are bank tellers who are also feminists? Once you’re counting people, the second number obviously can’t beat the first — and the error falls sharply. It’s the same trick that tames the base-rate fallacy.

Distrust the persuasive scenario. When a forecast, pitch or theory wins you over precisely because it’s so detailed and joins up so neatly, treat that as a warning. Every added specific is one more thing that has to come true.