The Wason Selection Task: Why Smart People Fail Simple Logic

A psychology experiment from 1966 presents you with four cards and a simple rule. You need to decide which cards to flip to find out if the rule is being broken. The task takes about 30 seconds to understand. Only about 10% of people get it right. The reason they fail is not stupidity — it is a systematic bias in how human reasoning works, one that shapes how scientists conduct research, how doctors interpret test results, and how everyone evaluates evidence for beliefs they already hold.

The Experiment and the Puzzle

Peter Cathcart Wason was a British cognitive psychologist at University College London who spent much of his career studying how humans reason about conditional statements — statements of the form "if P, then Q." In 1966, he published a short paper describing what he called the selection task, which was specifically designed to test whether people could correctly apply the logic of falsification.

Falsification — the process of trying to disprove a hypothesis rather than confirm it — is the cornerstone of scientific method as Karl Popper had described it. Wason wanted to know whether ordinary people, in the absence of scientific training, naturally think that way.

Source: Wason, P.C. (1966), "Reasoning," in B.M. Foss (ed.), New Horizons in Psychology, Penguin — the original publication of the selection task and the first report of participants' near-universal failure.

The basic version of the task is this. You are shown four cards lying on a table. Each card has a letter on one side and a number on the other. The four visible faces show:

E — K — 4 — 7

You are told the rule: "If a card has a vowel on one side, then it has an even number on the other side."

Your task: which cards do you need to turn over to determine whether the rule is being violated?

Take a moment to actually think about it before reading on.

The Correct Answer and Why Almost No One Gives It

The logically correct answer is to turn over E and 7.

Here is why.

The rule is a conditional: if vowel (P), then even number (Q). In formal logic, this conditional is violated in exactly one case: when P is true and Q is false — that is, when there is a vowel on one side and an odd number on the other. To test the rule, you need to check every card that could possibly produce this violation.

E: This card shows a vowel. You must turn it over to see whether the number on the back is even. If it is odd, the rule is broken.

K: This card shows a consonant. The rule says nothing about what number must accompany a consonant. Whether K has an even or odd number on the back, no violation occurs. You do not need to turn this card over.

4: This card shows an even number. The rule does not say "if even, then vowel" — it says "if vowel, then even." A card with an even number and a consonant on the back does not violate the rule. You do not need to turn this card over.

7: This card shows an odd number. If there is a vowel on the other side, the rule is violated. You must turn it over.

The correct answer is E and 7. In Wason's original experiments, fewer than 10% of participants gave this answer.

Source: Wason, P.C. & Johnson-Laird, P.N. (1972), Psychology of Reasoning: Structure and Content, Harvard University Press — the comprehensive account of early selection task research and the systematic documentation of failure rates.

The most common wrong answer is E and 4. Participants turn over the card that confirms the rule (the even number, to see if it has a vowel) rather than the card that could falsify it (the odd number, which if it has a vowel, breaks the rule). This error has a name: confirmation bias.

Source: Wason, P.C. (1960), "On the failure to eliminate hypotheses in a conceptual task," Quarterly Journal of Experimental Psychology 12(3), 129–140 — where Wason first coined and described confirmation bias, based on an earlier number-sequence experiment (the 2-4-6 task) that preceded the selection task.

What Confirmation Bias Actually Means

Confirmation bias — the tendency to search for, interpret, and recall information in a way that confirms what you already believe — is not the same as stubbornness or closed-mindedness. It is a deep-seated feature of how human cognition processes conditional statements.

When people see the rule "if vowel, then even number" and the card showing 4, they do not think "I should check whether this might break the rule." They think "the rule says vowels go with even numbers, and 4 is even, so this card might be relevant." They are thinking about what would confirm the rule, not what would refute it. The 7, which seems disconnected from the rule about vowels, does not leap out as the key falsifying card.

Think of it this way: if you believe a coin is biased toward heads, the natural thing to do is flip it several times and watch for heads. What falsification demands is that you think instead about what outcome would prove the coin fair — and then specifically look for that outcome. Most people do not naturally reframe hypotheses in terms of their potential refutation.

The practical consequence is significant. A doctor who believes a patient has condition X will tend to order tests that confirm X rather than tests that rule it out. A scientist who favors a theory will tend to design experiments that find supporting evidence rather than experiments that could expose a flaw. These are not moral failures — they are cognitive defaults.

The Thematic Content Discovery

Wason's original experiment used abstract symbols — letters and numbers — and produced the famous 10% success rate. But a series of follow-up experiments in the 1970s through 1990s showed that changing the content of the rule without changing its logical structure dramatically changed performance.

The most influential demonstration came from Richard Griggs and James Cox in 1982. They replaced the abstract rule with a social regulation: "If a person is drinking beer, then the person must be over 19 years old." The four cards showed: DRINKING BEER / DRINKING COKE / 22 YEARS OLD / 16 YEARS OLD.

Performance soared. Roughly 70–80% of participants correctly identified the relevant cards (DRINKING BEER and 16 YEARS OLD), compared to fewer than 10% on the abstract version.

Source: 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 — the foundational demonstration of content-dependent performance.

The logical structure is identical in both versions. The improvement must come from something about the social content. But what?

The Evolutionary Psychology Explanation

In the early 1990s, evolutionary psychologists Leda Cosmides and John Tooby proposed an influential explanation. They argued that humans did not evolve to reason about abstract conditional statements — we evolved to detect social cheaters. In cooperative social groups, the critical cognitive task was identifying people who take benefits without paying the required cost, and the selection task improves dramatically when it is framed as a cheater-detection problem.

Source: Cosmides, L. (1989), "The logic of social exchange: Has natural selection shaped how humans reason?" Cognition 31(3), 187–276 — the key evolutionary psychology paper presenting the social contract theory of the selection task.

Social contract theory — Cosmides and Tooby's account — holds that the human mind contains specialized reasoning algorithms for detecting violations of social exchanges, and that the selection task taps these algorithms only when framed as a social exchange problem. Abstract letters and numbers do not engage the cheater-detection module; drinking-age rules do, because they have the structure of "if you take this benefit, you must have met this requirement."

The theory is elegant and generated a large body of experimental support, but it has significant critics. Alternative explanations — including probabilistic accounts, domain familiarity, pragmatic reasoning schemas, and mental model theory — also explain the content effect without invoking evolved modules.

Source: Oaksford, M. & Chater, N. (1994), "A rational analysis of the selection task as optimal data selection," Psychological Review 101(4), 608–631 — a probabilistic alternative to the social contract account, arguing that the "correct" logical response is actually suboptimal from a Bayesian standpoint, and that people's behavior reflects rational information maximization.

Oaksford and Chater's point is worth dwelling on. They argued that if you assign realistic probabilities to the frequency of vowels, consonants, even numbers, and odd numbers in the real world, the logically "correct" falsifying strategy (turn over E and 7) actually yields less information than the "wrong" strategy (turn over E and 4). People choosing E and 4 may not be making an error of logic — they may be making a reasonable probabilistic inference that has the wrong label attached to it.

Results Across Different Framings

Rule Type	Example Content	Success Rate
Abstract letters/numbers	"If vowel, then even"	~10%
Familiar social rule	Drinking age example	~70–80%
Unfamiliar social rule	Obscure cultural rules	~10–30%
Arbitrary thematic	Non-social thematic content	~10–25%
Permission schema	"If action A, then must satisfy precondition B"	~50–65%
Obligation schema	"If condition C, then must take action D"	~40–60%

Source: Cheng, P.W. & Holyoak, K.J. (1985), "Pragmatic reasoning schemas," Cognitive Psychology 17(4), 391–416 — established the permission and obligation schema categories and their performance profiles.

The pattern reveals that abstract logical structure and content are not cleanly separable in human reasoning. The mind does not process "if P then Q" as a pure formal operation and then apply it to whatever content fills the slots. The content shapes the cognitive process used to evaluate the conditional.

Wason's Own Reaction to Participants

Wason was fascinated and reportedly delighted by how consistently people failed. He conducted extensive post-experiment interviews in which he would show participants the correct answer and explain why it was correct. Many participants agreed, verbally, that E and 7 was the right answer. Then Wason would ask them to do the task again, with a slightly different framing, and they would revert to E and 4.

This persistence of the error even after explanation suggests the bias is not simply ignorance of formal logic. It points to a processing style that does not automatically update based on verbal instruction — the same style that makes confirmation bias so difficult to correct through awareness alone.

The phenomenon is analogous to optical illusions: knowing that the two lines in the Müller-Lyer illusion are the same length does not make them look the same length. The perceptual system processes the visual information in a way that bypasses the verbal knowledge. Similarly, knowing about confirmation bias does not automatically produce falsifying reasoning.

Source: Wason, P.C. (1968), "Reasoning about a rule," Quarterly Journal of Experimental Psychology 20(3), 273–281 — includes the verbal explanations and post-experiment interviews that document the persistence of the error.

What the Experiment Tells Us About Science

The selection task has obvious implications for how scientific research is designed and evaluated. If researchers tend to design experiments that confirm existing theories rather than experiments that could refute them, the scientific literature will accumulate confirming evidence for wrong theories. This is not a hypothetical concern — it is a recognized problem in the published literature, particularly in fields where experimental design is complex and the same data can be analyzed in multiple ways.

The replication crisis in psychology and social science, which became a major conversation in the 2010s, is partly a consequence of confirmation bias operating at the level of research design. Researchers who believe a hypothesis is true tend to run experiments in ways that are more likely to confirm it, and they tend to seek and publish confirming results. Pre-registration — the practice of specifying in advance exactly what a study will measure and how it will be analyzed, before data collection — is the structural solution to this problem, forcing researchers to commit to a falsifying strategy rather than a confirming one.

Source: Open Science Collaboration (2015), "Estimating the replicability of psychological science," Science 349(6251), aac4716 — the large-scale replication study that found fewer than half of published psychology findings replicated, partly attributable to confirmation bias in original research design.

The Mundane Version That Runs Your Life

The selection task's lesson is most important in everyday contexts that never announce themselves as logic problems. When a hiring manager believes a candidate is qualified, they tend to ask questions that confirm qualifications rather than probe weaknesses. When a voter believes a politician is corrupt, they interpret ambiguous evidence as confirming corruption. When an investor believes a stock will rise, they seek out analyst reports that support the bullish case.

None of these situations present themselves as "if P, then Q" problems requiring card-turning logic. They feel like ordinary judgment calls. But the underlying cognitive operation is identical: the mind reaches for confirming evidence first, and it requires deliberate effort — often failing even then — to search instead for evidence that would force reconsideration.

This is what makes the Wason selection task so enduring as an experimental tool. It is not really a test of logic. It is a clean, reproducible demonstration of a cognitive style that shapes nearly every belief a human being forms and nearly every piece of evidence they bother to look at.

The Takeaway

Peter Wason's four-card puzzle has been run in hundreds of variations on tens of thousands of participants across sixty years, and the core finding has not changed: most people seek confirming evidence for their beliefs rather than evidence that could falsify them, and this tendency is so robust that explaining the correct logic often fails to correct the error in the next attempt. The content of the rule matters enormously — social cheating problems unlock reasoning abilities that abstract logic problems cannot — which suggests the human mind is not a general-purpose logic engine but a collection of specialized tools shaped by the problems our ancestors regularly faced. Knowing this does not make you immune to it. But it is a useful fact to carry into any situation where the evidence you are considering already agrees with what you expected.

---

Read next:

• The Dunning-Kruger Original Experiment
• The Loftus Misinformation Experiment
• The Iowa Gambling Task

---

Q: What is the correct answer to the Wason selection task and why?
A: The correct cards to turn over are E (the vowel) and 7 (the odd number). The rule — "if vowel, then even number" — is violated only when a vowel has an odd number on its back. The E card might reveal such a violation. The 7 card, if it has a vowel behind it, is a direct violation. The K card is irrelevant because the rule makes no claim about consonants. The 4 card is irrelevant because finding a consonant behind an even number does not break the rule.

Q: Why do people get the Wason task wrong even after being told the correct answer?
A: Knowing the logical principle verbally does not automatically change the cognitive process used during the task. Confirmation bias appears to operate at a processing level that is at least partially independent of explicit verbal knowledge, similar to how knowing an optical illusion is an illusion does not change what you see.

Q: Does IQ score predict performance on the Wason task?
A: Higher IQ scores are weakly correlated with better performance on abstract versions of the task, but the correlation is modest. Highly intelligent people show the same content-dependent patterns as less intelligent people — they do much better on social-rule versions and poorly on abstract versions — suggesting the effect is not primarily about general intelligence.

Q: What is the social contract theory of the Wason task?
A: Evolutionary psychologists Leda Cosmides and John Tooby proposed that humans have evolved cognitive mechanisms specifically for detecting social cheaters — people who take benefits without fulfilling obligations. The selection task improves dramatically when framed as a cheater-detection problem because it engages these evolved mechanisms, while abstract letter-number versions do not.

Q: Has the experiment been used in real-world training programs?
A: Yes, versions of the task are used in critical thinking courses, medical education, and legal reasoning training to illustrate confirmation bias. The typical finding is that people can learn to perform better on specific task types through practice, but transfer to novel domains is limited — improving on drinking-age problems does not reliably improve performance on abstract letter-number problems.

Q: What did Wason himself conclude about human rationality?
A: Wason was skeptical of strong claims about human rationality. He believed the experiment showed that humans do not naturally reason in accordance with formal logic, and he was particularly struck by the fact that the error persisted even after participants were shown the correct answer and agreed with the explanation. He saw this as evidence that human cognition is governed by processes that are at best partially accessible to conscious reasoning.