The conjunction fallacy, and why more detail feels likelier

An everyday example

A news story describes a small, struggling shop closing down after a well-known chain store opened nearby. Asked which is more likely, "the shop closed because of financial trouble" or "the shop closed because of financial trouble caused specifically by the new chain store," many people pick the second, more detailed explanation, because it fits the story so neatly. The second explanation is a narrower version of the first: every shop that closed for that specific reason also closed from financial trouble, so it cannot be the more probable of the two, no matter how well the specific story fits.

A more detailed story often feels more convincing than a plainer one, even though adding detail can only narrow how many ways it can be true, never widen it.

The classic experiment

Tversky and Kahneman's 1983 paper gave participants a short character sketch of a young, unmarried woman, outspoken and deeply concerned with issues of social justice, who had studied philosophy at university, and then asked them to compare the probability of several statements about her, including that she works as a bank teller, and that she works as a bank teller and is also active in the feminist movement. Logically, the second statement can never be more probable than the first, since everyone who fits the second description also fits the first: being a feminist bank teller is one way of being a bank teller, not a separate possibility alongside it. In the study, a large majority of participants rated the more detailed, two-part description as more probable than the plain one-part description, a direct violation of a basic rule of probability that a combination of two conditions cannot exceed either condition alone.

The character sketch worked because it matched a stereotype of political engagement far better than it matched a stereotype of banking, so "feminist bank teller" felt like a more representative, better-fitting story than plain "bank teller" did, even though it is a strictly narrower, and therefore no more probable, possibility.

Does it replicate?

Replication grade: Mixed: the effect is reliable under the original wording; framing as frequencies shrinks it a great deal

The basic finding, that a detailed, representative-sounding conjunction gets rated as more probable than one of its own broader components, has been reproduced across many versions of the task since 1983, with different character sketches, different professions and different combinations of traits. It is one of the most widely cited demonstrations in the study of probability judgement.

Ralph Hertwig and Gerd Gigerenzer's 1999 reconsideration showed the size of the effect depends heavily on how the question is asked. When the same kind of problem was reframed to ask how many out of a stated group of 200 women fit each description, rather than asking for a single-case probability judgement, the violation of the conjunction rule largely disappeared. They argued that part of the original effect comes from ordinary conversational habits: outside a maths classroom, the word "probability" is often read as something closer to plausibility or fit, and some of what looks like a logical error may be a reasonable answer to a differently understood question. That does not erase the finding under the original wording, which still shows the error reliably; it shows the size of the error is sensitive to exactly how the comparison is posed.

Why the smaller circle cannot be the bigger probability

The diagram on this page is an illustration, not measured data. It draws two nested circles, everyone who fits a broad description and, entirely inside it, everyone who also fits a narrower, more detailed one. Because the inner group is entirely contained within the outer one, it can only be smaller or equal to it, never larger, however much better its story seems to fit.

Try it: three scenarios

Three original scenarios, with invented names, built the same way as the classic study: a detailed description, then a choice between a broad possibility and a narrower, more specific one that is entirely contained within it.

How to catch it

The fallacy is easiest to catch by checking containment directly, rather than by trusting how convincing a story feels.

• Ask directly: is everyone who fits the detailed description also a member of the broader group? If so, the detailed one cannot be more probable.
• Rewrite the comparison as two circles, one inside the other, before judging which is more likely.
• Notice when a more detailed version of an explanation feels more convincing purely because it paints a fuller picture, not because it is backed by more evidence.
• Convert the comparison into counts: out of a hundred cases fitting the broad description, how many would also fit the narrow one; the narrow count can never exceed the broad one.
• Be suspicious of any story that adds a specific, plausible-sounding extra detail without adding any actual evidence for that detail.