Framing of logic
Logically identical problems get solved or flubbed depending on how they are dressed.
What it means
The framing of logic refers to how the surface content, context, and presentation of a logically constant problem dramatically change reasoning performance, even though the formal structure is unchanged. The same conditional is solved far better when cast as a familiar social-permission rule than as an abstract letters-and-numbers task, and syllogisms are judged more accurately when their conclusions are neutral than when belief intrudes. This content-dependence is a foundational finding: human reasoning is not a content-free logic engine but is cued, scaffolded, or derailed by what the problem is about. Explanations range from pragmatic reasoning schemas and evolved content-specific mechanisms to the differential availability of counterexamples and mental models. Recognizing it means that to improve reasoning one can re-frame a problem into a form the mind handles well, and that persuasive framing can smuggle in error.
The original demonstration
The phenomenon was first isolated with Peter Wason's selection task, introduced in the late 1960s. Participants see four cards, each carrying a letter on one side and a number on the other, together with a rule of the form "if a card has a vowel on one side, it has an even number on the other." They must name only the cards that need turning to test the rule. The logically correct answer is the vowel card and the odd-number card, the two that could falsify the rule; yet only around a tenth of educated adults select these, most turning the vowel and the even number instead. The result was striking precisely because the task is formally trivial. Within a few years researchers found that replacing the abstract letters and numbers with realistic material could transform performance. An early study cast the same rule as a postal regulation about sealed envelopes requiring higher postage, and many participants who had failed the abstract version now succeeded. That contrast, identical logical form but very different success, is the founding observation behind the framing of logic.
What the evidence shows
The early optimism that realism itself unlocks logic did not survive replication. When Richard Griggs and James Cox revisited the thematic-materials effect in 1982, they could not reproduce the postal-regulation result or a widely cited transport-rule result; concrete, sensible content on its own left error rates close to the abstract baseline. What did reliably help was content the reasoner had direct experience of. A rule about being old enough to drink alcohol in a bar produced correct selections in roughly seventy percent of participants. This first suggested memory of specific rules, but Patricia Cheng and Keith Holyoak showed in 1985 that facilitation does not require prior familiarity at all. An unfamiliar rule invented for the experiment, but phrased as a permission of the form "to take an action one must first satisfy a condition," was enough to lift performance, implying that a general permission structure rather than the particular memory was doing the work. A separate and more robust literature concerns syllogisms. Evans, Barston and Pollard reported in 1983 that people endorse conclusions they find believable and reject ones they do not, largely regardless of whether the conclusion actually follows; this belief bias is stronger for invalid arguments, where belief and logic conflict. Klauer and colleagues later modelled the same kind of data and confirmed that both a genuine logic component and a separate belief-driven response bias are needed to fit it.
Why it happens
No single mechanism has won the field. Pragmatic reasoning schemas hold that people possess generalized rule-sets tied to goals such as permission, obligation and causation, and that a problem activating such a schema is handled with the schema's built-in checking procedures rather than by formal logic. Social-contract theory, advanced by Leda Cosmides in 1989, locates the effect in an evolved mechanism for policing exchanges: the mind is claimed to be especially good at spotting someone who takes a benefit without paying the cost, which happens to coincide with the logically correct selection. Both accounts were challenged by relevance theory, which argues that no special reasoning machinery is required. On this view the wording of a rule alters which cases come to mind as relevant, and once the falsifying cases are made salient by the phrasing, the correct cards are simply the obvious ones to check. A fourth tradition, mental-model theory, attributes belief bias and content effects to how easily counterexamples are constructed and searched: a believable conclusion discourages the effortful hunt for a model in which the premises hold but the conclusion fails. These explanations are not fully separable, and much of the dispute turns on whether content genuinely improves reasoning or merely changes what looks worth reasoning about.
Where it shows up
The pattern surfaces wherever the same logical demand is dressed in different content. Regulatory and safety rules, such as who may enter a restricted area or which transactions require a signature, are exactly the permission-and-obligation conditionals that people tend to check accurately, which is one reason procedures are often written as permissions rather than abstract conditions. Belief bias appears whenever arguments are evaluated by people with a stake in the conclusion: a policy brief, a peer review, or an instruction whose conclusion aligns with prior conviction tends to be judged sound, while an equally strong argument for an unwelcome conclusion is scrutinized for flaws. Persuasion exploits the same lever in the other direction, wrapping a shaky inference in agreeable, familiar content so that it passes unexamined. The practical upshot cuts both ways. A reasoning task can be redesigned into a form the mind handles well, by recasting an abstract check as a concrete permission or by stripping emotionally loaded content out of an argument before its structure is assessed. The same sensitivity, though, means that framing can be used to launder error, making a fallacy feel like plain common sense.
Limits and caveats
The framing of logic is an umbrella over several distinct phenomena that should not be merged. Deontic facilitation in the selection task, belief bias in syllogisms, and content effects on conditional inference have different signatures and probably different causes, so a mechanism that explains one need not explain the others. The facilitation is also narrower than it can appear. Getting the permission version right does not show that people have become better at logic; on relevance-theory and pragmatic accounts they may be following a social rule or a salience cue that only coincides with the logical answer, and they frequently cannot transfer the success to a structurally identical abstract problem. Belief bias, for its part, is moderated by individual differences and by time and motivation: readers higher in cognitive ability, or prompted to deliberate, resist it more, though few eliminate it entirely. And the direction of an effect depends on details of wording that are hard to predict in advance, which is why realism sometimes helps, sometimes does nothing, and occasionally misleads. The safe generalization is modest. What a problem is about reliably changes how well it is reasoned through, but not always in the direction, or for the reasons, that intuition suggests.
Examples
People who fail an abstract version of the Wason task solve its logical twin instantly when it reads 'if someone is drinking beer, they must be over 21.'
Asked to check whether a rule like "anyone served alcohol must be over eighteen" is being obeyed, most people correctly turn over both the underage customer and the person drinking beer; the same people, given a structurally identical rule about vowels and even numbers on cards, mostly fail.
Presented with a syllogism whose conclusion is believable but does not in fact follow from the premises, many readers endorse it as valid; when the identical invalid structure yields an unbelievable conclusion, they correctly reject it, so accuracy tracks believability rather than logic.
A troubleshooting checklist written as an abstract if-then condition produces frequent oversights, but rewritten as a permission a technician must verify before proceeding, the very same check is completed reliably.
Dressing a reasoning problem in vivid, realistic detail does not by itself improve performance: a concrete rule about towns and modes of travel, though familiar, can leave error rates almost as high as the abstract version.
First described in Wason; Cheng & Holyoak (pragmatic reasoning schemas, 1985).
Key references
- Wason, P. C. (1968). Reasoning about a rule. Quarterly Journal of Experimental Psychology, 20(3), 273-281. doi.org/10.1080/14640746808400161
- Johnson-Laird, P. N., Legrenzi, P., & Legrenzi, M. S. (1972). Reasoning and a sense of reality. British Journal of Psychology, 63(3), 395-400. doi.org/10.1111/j.2044-8295.1972.tb01287.x
- Griggs, R. A., & Cox, J. R. (1982). The elusive thematic-materials effect in Wason's selection task. British Journal of Psychology, 73(3), 407-420. doi.org/10.1111/j.2044-8295.1982.tb01823.x
- Evans, J. St. B. T., Barston, J. L., & Pollard, P. (1983). On the conflict between logic and belief in syllogistic reasoning. Memory & Cognition, 11(3), 295-306. doi.org/10.3758/BF03196976
- Cheng, P. W., & Holyoak, K. J. (1985). Pragmatic reasoning schemas. Cognitive Psychology, 17(4), 391-416. doi.org/10.1016/0010-0285(85)90014-3
- Cosmides, L. (1989). The logic of social exchange: Has natural selection shaped how humans reason? Studies with the Wason selection task. Cognition, 31(3), 187-276. doi.org/10.1016/0010-0277(89)90023-1
- Sperber, D., Cara, F., & Girotto, V. (1995). Relevance theory explains the selection task. Cognition, 57(1), 31-95. doi.org/10.1016/0010-0277(95)00666-M
- Klauer, K. C., Musch, J., & Naumer, B. (2000). On belief bias in syllogistic reasoning. Psychological Review, 107(4), 852-884. doi.org/10.1037/0033-295X.107.4.852