Conjunction fallacy
Also known as: Linda problem
A vivid, specific story can seem likelier than the plain fact it contains.
What it means
The conjunction fallacy is the error of judging the joint occurrence of two events as more probable than one of those events alone — a logical impossibility, since adding a condition can only narrow the set of cases and therefore can never raise probability. It arises because people judge by representativeness rather than by the rules of probability: a richly detailed scenario matches a stereotype better than a sparse one, so it 'feels' more likely even though it is strictly less probable. The famous Linda problem is the canonical demonstration, but the effect generalizes to forecasting, where adding plausible-sounding detail to a prediction makes it seem more credible while making it mathematically less likely. A live debate concerns how much of the effect reflects genuine probabilistic confusion versus pragmatic misreading of the word 'probable' or of 'bank teller' as implying 'and not a feminist'; framing the task in frequencies reduces but does not eliminate it. In practice it warns against being persuaded by detailed, coherent narratives in intelligence analysis, planning, and risk assessment, where specificity masquerades as likelihood.
The original demonstration
Tversky and Kahneman introduced the effect in 1983 with the story of Linda, a bright, outspoken philosophy graduate concerned with justice. Asked to rank statements, most respondents judged "Linda is a bank teller and active in the feminist movement" as more probable than "Linda is a bank teller" alone, though the first is a subset of the second. Around 85 percent committed the error, including graduate students trained in statistics. The pattern was not confined to Linda: the authors reproduced it in medical prognosis, sports and political forecasting, and even estimates of word frequency, and it survived when both options sat side by side rather than being split between separate groups of readers.
Why the conjunction wins
The gloss attributes the effect to representativeness, but look closer and the driving quantity seems to be fit between the evidence and the added detail. Katya Tentori and colleagues argue people are not estimating probability at all but inductive confirmation, how strongly the description supports each hypothesis. "Feminist" is exactly what Linda's profile predicts, so tacking it on raises the story's perceived confirmation even as it lowers its probability, and the two pull in opposite directions. Consistent with this, the fallacy grows when the added conjunct is well supported by the vignette and fades when it is irrelevant or contradicts it. The mind answers a question about coherence and mistakes the answer for one about odds.
Is it a genuine fallacy?
Sceptics argue the error is partly an artefact of language. Ralph Hertwig and Gerd Gigerenzer noted that "probable" has everyday senses closer to "plausible" or "telling," and that "and" can be read as a set union, so a respondent who ranks by those meanings is not violating logic. Asking how many of 100 women like Linda are bank tellers lowers error rates sharply. But the pragmatic story is only partial. In a 2001 adversarial collaboration, Mellers, Hertwig and Kahneman agreed that frequency wording alone did not remove the effect; it disappeared only when filler items were dropped and the phrasing was disambiguated. Something beyond mere misreading of the question remains.
What shrinks it
The effect is real but movable. Charness, Karni and Levin found that a small payment for the correct answer cut violation rates, and letting people discuss the problem in threes cut them further, suggesting many respondents can recover the logic once it is worth the effort. A 2023 meta-analysis put the debiasing effect of incentives at a modest but reliable size, larger where baseline accuracy was low. The practical lesson for forecasting, intelligence and planning is not that people are hopeless but that vivid, well-fitting detail quietly inflates perceived likelihood. Stripping a scenario back to its bare claim, or attaching a cost to being wrong, restores some of the discipline that the story stripped away.
Examples
Told about Linda, a justice-minded philosophy graduate, most rate 'bank teller and feminist' as more likely than 'bank teller.'
Forecasters rate 'a recession next year triggered by an energy shock' as likelier than 'a recession next year' — the added cause makes the story vivid, and strictly less probable.
Travellers will pay more for insurance against death on a flight due to terrorism than for insurance against death on a flight from any cause — which already includes terrorism.
A clinician rates "the patient has a cough and lung cancer" as more likely than "the patient has a cough," because the grim, specific diagnosis fits a worried mental picture better than the bare symptom.
A bettor takes short odds on "the favourite wins after conceding the first goal" over the plain "the favourite wins," because the comeback narrative is vivid, though it can only be the rarer outcome.
First described in Tversky & Kahneman (1983).
Key references
- Yechiam, E., & Zeif, D. (2023). The effect of incentivization on the conjunction fallacy in judgments: A meta-analysis. Psychological Research, 87(8), 2336-2344. doi.org/10.1007/s00426-023-01837-5
- Tentori, K., Crupi, V., & Russo, S. (2013). On the determinants of the conjunction fallacy: Probability versus inductive confirmation. Journal of Experimental Psychology: General, 142(1), 235-255. doi.org/10.1037/a0028770
- Charness, G., Karni, E., & Levin, D. (2010). On the conjunction fallacy in probability judgment: New experimental evidence regarding Linda. Games and Economic Behavior, 68(2), 551-556. doi.org/10.1016/j.geb.2009.09.003
- Mellers, B., Hertwig, R., & Kahneman, D. (2001). Do frequency representations eliminate conjunction effects? An exercise in adversarial collaboration. Psychological Science, 12(4), 269-275. doi.org/10.1111/1467-9280.00350
- Hertwig, R., & Gigerenzer, G. (1999). The "conjunction fallacy" revisited: How intelligent inferences look like reasoning errors. Journal of Behavioral Decision Making, 12(4), 275-305. onlinelibrary.wiley.com/doi/abs/10.1002/(SICI)1099-0771(199912)12:4%3C275::AID-BDM323%3E3.0.CO;2-M
- Tversky, A., & Kahneman, D. (1983). Extensional versus intuitive reasoning: The conjunction fallacy in probability judgment. Psychological Review, 90(4), 293-315. doi.org/10.1037/0033-295X.90.4.293