What is logical reasoning?

The Wason selection task

The best-known logic puzzle in psychology is the Wason selection task, developed in 1966. You see four cards and must decide which ones to turn over to test a rule linking the letter on one side to the number on the other (APA Dictionary: Wason selection task). Suppose the cards show E, K, 4 and 7, and the rule is "if a card has a vowel on one side, it has an even number on the other." The right answer is E and 7: the 7 is the only number that could break the rule.

Most people fail the task by not checking the cards that could break the rule (APA Dictionary: Wason selection task). On the standard abstract version, only around 10% of people choose the correct pair (the P card and the not-Q card); most choose P and Q, or P alone (Sperber, Cara & Girotto, 1995). Wason (1968) traced the difficulty to a mental set that expects sentences to match states of affairs, which blocks the contrapositive inference (Wason, 1968).

The task has been studied heavily: a 2018 analysis pooled the results of 228 experiments (Ragni, Kola & Johnson-Laird, 2018). Content matters less than people once hoped. Griggs and Cox failed to replicate an earlier report that realistic content improves performance, but found improvement for a rule taken from participants' own past experience (Griggs & Cox, 1982).

Belief bias: when believable beats valid

Belief bias is the tendency to accept conclusions because they are believable rather than because they are logically valid (APA Dictionary: belief bias). It is usually measured with syllogisms in which a conclusion's believability conflicts with its validity (APA Dictionary: belief bias). Try this one: all flowers need water; roses need water; so roses are flowers. The conclusion is true, but it does not follow. Plenty of things need water without being flowers.

In a classic study, Evans, Barston and Pollard found that people accepted more valid than invalid conclusions and more believable than unbelievable ones, and that belief bias was stronger on invalid syllogisms (Evans, Barston & Pollard, 1983) (Davies, 2007 (Stanford course notes)). In other words, a believable conclusion is most dangerous exactly when the logic behind it is broken.

Is logical reasoning part of IQ?

Partly. Logic items appear on many ability tests, and dual-process research links abstract reasoning to general intelligence (Evans, 2003). But a high score does not protect you from every reasoning error. Across 7 studies, many classic thinking biases, including framing, anchoring, base-rate neglect, myside bias and the sunk-cost effect, were uncorrelated with cognitive ability (Stanovich & West, 2008). Cognitive ability did correlate with avoiding some biases, including belief bias and matching bias on the selection task (Stanovich & West, 2008). So the logic that IQ tests sample overlaps with, but is not the same as, thinking well in daily life. The question can smart people be irrational? looks at that gap.

How logical reasoning is measured

Research batteries use syllogism and inference tests like the ETS kit's, and clinical batteries use tasks such as Analysis-Synthesis (McGrew, 2023). Employment tests fold logic into short timed mixes. The CCAT has 50 questions in 15 minutes, in three kinds: verbal, math and logic, and spatial reasoning (Criteria: CCAT prep). The PI Cognitive Assessment is a 50-question, 12-minute test covering simple math, basic logic and language comprehension (Predictive Index: Cognitive Assessment). The CCAT page explains how those scores are reported.

Nonverbal IQ tests usually test logic through figures rather than sentences: an analogy (A is to B as C is to what?) or a grid whose rows follow a rule. The IQ Octopus test includes a Logical Reasoning section of 3 items out of 25, and the whole test produces an estimated IQ, not a separate logic score. The figure analogies guide shows how those items are built.

Can you improve logical reasoning?

On a specific task, yes, and sometimes quickly. When people got feedback on repeated selection tasks, they switched from picking cards that confirm the rule to picking possible counterexamples (Ragni, Kola & Johnson-Laird, 2018). Retaking any ability test also helps: average retest gains were 0.33 SD at the second sitting and 0.50 by the third, with no further gains after the third (Scharfen, Peters & Holling, 2018).

Broader claims are weaker. A 2016 meta-analysis found no convincing evidence that working memory training improves far-transfer measures such as nonverbal ability when compared with a treated control group (Melby-Lervåg et al., 2016), and a review of several meta-analyses found minimal effects of cognitive training on domain-general cognitive skills (Sala & Gobet, 2019). The most consistent route found so far is education: an additional year of schooling raises cognitive ability by approximately 1 to 5 IQ points (Ritchie & Tucker-Drob, 2018). Practicing logic puzzles makes you better at logic puzzles; it is not a shortcut to a higher general ability.

Try it

The quickest way to see where you stand is to work a few items under time pressure. Take the free logical reasoning test for ten timed analogies and grids, then browse the analogy puzzles for untimed practice with worked answers.