Conjunction reasoning
Inference about 'and' — combining conditions, which can only narrow the possibilities.
What it means
Conjunction reasoning concerns judgments and inferences about co-occurring conditions joined by 'and,' where logically the probability of a conjunction can never exceed that of either conjunct alone. People nonetheless violate this when a detailed, representative scenario feels more plausible than its broader components, the basis of the conjunction fallacy. The error is sharpest when added specifics increase a story's coherence or fit to a stereotype, trading logical extension for narrative plausibility. Framing in frequencies, clarifying the set relations, or prompting analytic processing reduces the violation. Studying conjunction reasoning reveals a deep tension between how we evaluate explanatory fit and how probability actually behaves.
The rule it rests on
The conjunction rule is a theorem, not an empirical claim. Because every case that is both A and B is also a case of A, the set of A-and-B outcomes sits inside the set of A outcomes and can be no larger, whatever probabilities you assign. That is why a violation counts as a logical error rather than a bad estimate. Researchers test it two ways. A between-subjects design shows different people either the conjunction or the single conjunct; a within-subjects, or direct, design puts both in front of the same person. The fallacy survives even the within-subjects version, where both options appear side by side, which shows that people are not merely failing to notice the subset relation.
Why people do it is contested
Tversky and Kahneman blamed representativeness: the conjunction resembles the stereotype more closely, and resemblance substitutes for probability. Later accounts diverge sharply. Hertwig and Gigerenzer argue much of the effect is conversational, since in ordinary use 'probable' can mean plausible or telling, and 'Linda is a bank teller' invites the reading 'and not a feminist' — so the ranking becomes a sensible answer to an ambiguous question. Tentori, Crupi and Russo make a different claim: the real driver is confirmation, not probability, because people favour the conjunct that the description most strongly supports, and this predicts the fallacy better than similarity does. Costello and Watts instead model judgments as unbiased estimates corrupted by random noise. No account has settled the matter.
How robust it is, and what shrinks it
The phenomenon replicates readily; the Linda problem draws violation rates near 80 to 90 percent. Its size, though, depends heavily on wording. Asking in frequencies — 'of 100 people like Linda, how many are…' — can drop violations from a clear majority to a small minority, because it makes the set inclusion visible. Paying for correct answers helps only a little: a 2023 meta-analysis of some 3,000 participants put the incentive effect at about d = 0.19, a nudge rather than a cure. Small-group discussion reduces it further. Clearer phrasing, separated ratings, and prompts toward analytic rather than intuitive processing all lower the rate, which is why the effect is best described as reliable in occurrence but fragile in form.
Where it bites, and how to blunt it
Detailed forecasts are its natural habitat. An intelligence estimate or a business plan that names many specific conditions reads as more convincing precisely because it says more, yet every added clause is one more way to be wrong; piling on plausible detail can raise judged probability while lowering the true probability. The same trap appears in diagnosis, where one vivid extra symptom makes a syndrome feel likelier, and in the courtroom, where a richly specified account of events persuades better than a bare one. The defenses are mechanical rather than clever: rate each condition on its own before combining them, state chances as frequencies, and treat any story whose appeal grows with its specificity as a warning sign, not a selling point.
Examples
People rate 'is a vegetarian and a yoga teacher' as more likely than simply 'is a vegetarian,' because the fuller portrait coheres — though it cannot be more probable.
A suspect who 'broke in through the window and fled in a stolen car' makes a tighter, more satisfying story than one who simply 'broke in' — and, because it says more, a longer shot.
A pitch promising 'a million users and profitability by year three' reads as more credible than 'a million users,' though every extra condition is one more way for the plan to fall short.
A forecaster who predicts 'a cold snap and a grid failure next week' seems more likely to be right than one predicting just 'a grid failure' — though pairing the claims only makes them rarer.
Voters rate 'the incumbent loses and turnout hits a record' as more probable than 'the incumbent loses,' because the fuller storyline feels like one coherent account of the same election.
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
- Costello, F., & Watts, P. (2014). Surprisingly rational: Probability theory plus noise explains biases in judgment. Psychological Review, 121(3), 463–480. doi.org/10.1037/a0037010
- 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
- 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. doi.org/10.1002/(SICI)1099-0771(199912)12:4<275::AID-BDM323>3.0.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