What Does an IQ Test Actually Measure — and What It Misses

An IQ test measures your performance on a short set of specific mental tasks. On the current professional adult test — the WAIS-5 — those tasks cover five abilities: verbal comprehension, visual-spatial skill, fluid reasoning, working memory and processing speed. That is the whole list. So when people ask what an IQ test actually measures, the honest answer is narrower than the word "intelligence" suggests. It is about 45 minutes of puzzles, definitions and memory drills, scored against how other adults your age do on the same tasks.

What the score leaves out is just as well documented: creativity, persistence, practical skill, emotional regulation. And some of the best-known claims on both sides have been overturned in the past few years — including one famous study that was quietly retracted.

What an IQ test actually measures: five abilities, ten tasks

The WAIS-5, released by Pearson in 2024, replaced the WAIS-IV of 2008. It builds a Full Scale IQ from five index scores, and each index rests on two core tasks. Here is what you actually do:

• Verbal comprehension. Similarities: say how two things are alike — how are a chair and a table alike? Vocabulary: explain what words mean.
• Visual-spatial skill. Block Design: recreate a printed pattern with red-and-white blocks, against the clock. Visual Puzzles: pick the three pieces that would combine into a shown shape.
• Fluid reasoning. Matrix Reasoning: choose the piece that completes a visual pattern. Figure Weights: work out which shapes will make a scale balance.
• Working memory. Digit tasks: listen to strings of numbers and repeat them back in order, or track a stream of digits and recall the most recent ones.
• Processing speed. Coding: using a key that pairs numbers with symbols, copy the right symbol under each number as fast as you can. Symbol Search: scan rows of symbols for a target, against a time limit.

That is the entire machine. Pearson says the seven subtests behind the Full Scale IQ take about 45 minutes, and all ten primary subtests take about an hour. The children's version, the WISC-V, has used the same five areas since 2014. The older WAIS-IV had only four indexes; the fifth edition split its "perceptual reasoning" score into visual-spatial and fluid reasoning. An IQ score is a summary of small, mostly timed tasks like these — not a reading taken from some hidden quantity.

The score is a comparison, with a built-in margin of error

Your Full Scale IQ does not measure an amount of anything. It ranks your combined performance against a norm group of people your age who took the same tasks. Two things follow from that.

First, the score has a known wobble. The WAIS-IV technical manual put the average standard error of measurement for the full score at about 2 points — the shift you would expect between sittings. At 95 percent confidence, that is a band of roughly plus or minus 4 points. Pearson's own score reports print confidence intervals, not single numbers, for exactly this reason.

Second, the score is a sample of one day. It records how you performed in one sitting, and nothing about the test can tell those 45 minutes apart from your average.

How much of school and work the score really explains

A 2015 meta-analysis in the journal Intelligence pooled decades of research on IQ and school grades. The corrected correlation came out at.54. That is one of the strongest links in psychology — and it still means intelligence accounts for under a third of the differences in grades. Most of what separates students is something else.

For job performance, the picture recently changed. The classic 1998 estimate put the corrected correlation at.51, which made cognitive ability the best-known predictor of work performance. A 2022 reanalysis in the Journal of Applied Psychology argued the old statistical corrections were overdone and re-estimated it at.31. Researchers are still arguing over which figure is right. Either way, most of what makes someone good at a job sits outside the score.

Creativity: nothing on the test rewards a new idea

Look back at the ten tasks. Every one converges on an answer the examiner already has. None asks you to produce something original. Creativity researchers measure that separately, with divergent-thinking tests — think of as many unusual uses for a brick as you can, and get scored on novelty.

How related are the two? A 2005 meta-analysis covering more than 45,000 people found a correlation of.17 between IQ and creativity-test scores — the author called the relationship negligible. A 2021 update across 112 studies put it at.25. Real, but small.

What about the popular claim that IQ helps creativity only up to about 120? The evidence has turned against it. A 2013 Austrian study of 297 adults found the supposed threshold moved — roughly 85, 100 or 120 — depending on how creative potential was defined, and found no threshold at all for actual creative achievement. Higher intelligence kept helping. A 2020 reanalysis in two larger samples found no threshold anywhere and recommended abandoning the idea altogether.

The famous motivation finding was retracted in 2025

For years, the standard citation on motivation was a 2011 analysis in PNAS of about 2,000 participants. It reported that offering material incentives raised IQ scores by 0.64 standard deviations — close to ten points — with bigger gains for lower scorers.

In May 2025, the authors retracted the paper. One study in the analysis, responsible for 24 percent of the participants, had itself been retracted. Without it, the effect dropped to 0.47. After adjusting for publication bias, it fell to 0.16, with a confidence interval that included zero. A 2022 study in Intelligence had already concluded that effort changes scores only in a minor way.

The deeper point survives without the big number. A test records what you did, not why. It cannot tell a mind that could not solve the matrix from one that stopped caring at minute forty. But nobody should quote ten points anymore.

Persistence: what Terman's high-IQ children revealed

Starting in the 1920s, Lewis Terman followed a large group of California children selected for very high IQs. At the 25-year follow-up, his team compared the 150 most successful men in the group with the 150 least successful. Terman's own verdict, published in 1947: "intellect and achievement are far from perfectly correlated." Accounts of the follow-up point to perseverance and self-confidence as separators — and also to family background, since the less successful group came from less advantaged homes.

Modern data agrees that character pulls real weight. A 2009 meta-analysis of more than 70,000 students found that once earlier school results were controlled, conscientiousness added as much to predicting university performance as intelligence did. No IQ subtest measures whether you keep showing up.

Practical intelligence: expert at the racetrack, average on paper

In 1986, researchers studied expert racetrack handicappers — men who priced horses for a living. The American Psychological Association's 1996 task-force report on intelligence summed up the result: the skilled handicappers implicitly used "a highly complex interactive model with as many as seven variables," and the ability to do so was unrelated to IQ. The study was small and has published critics, but the pattern recurs elsewhere. A 1985 study of child street vendors in Recife, Brazil compared the arithmetic they did inside real sales with the same operations posed as school-style problems, and documented street strategies that looked nothing like written school arithmetic.

IQ tasks strip away real-world context on purpose — that is what makes scores comparable across people. The cost is that skill embedded in a real setting goes uncounted.

Emotional regulation lives on a different test

No Wechsler task asks you to read a face, defuse an argument, or keep your composure under pressure. Those abilities are measured by separate instruments — the best known is the MSCEIT, which includes tasks specifically on managing emotions.

And they are not just IQ in disguise. A 2020 meta-analysis in Psychological Bulletin, covering 42,529 students, found that ability-based emotional intelligence correlated.24 with academic performance — and still added predictive power after intelligence and personality were both accounted for. In that analysis it ranked third among predictors, behind intelligence and conscientiousness. Related to IQ, then, but not contained in it.

How to read the number

An IQ score is a careful measure of a narrow, real thing: reasoning with words and patterns, holding numbers in mind, deciding quickly. Within that band, it predicts school and work outcomes better than any other single psychological measure — and still leaves most of both unexplained. Outside the band, it is silent. The new idea, the refusal to quit, the market-stall arithmetic, the cool head in an argument: none of it appears on the score sheet. Not because those things cannot be measured, but because nobody put them on this test. The test's own publisher prints every score with a confidence interval and a date. That is the right way to read yours — an estimate, of defined abilities, from one sitting.