How do IQ tests work?

IQ tests present a series of reasoning problems — pattern matrices, verbal analogies, memory and speed tasks — and compare your raw performance to a large norming sample of people your age. Your position in that distribution is converted to a score with a mean of 100 and a standard deviation of 15.

What do IQ tests measure?

IQ tests measure a cluster of cognitive abilities: abstract and fluid reasoning, verbal comprehension, working memory and processing speed. They do not measure creativity, emotional intelligence, motivation or character.

How are IQ tests scored?

Modern tests use deviation scoring. Your raw score (number correct, sometimes weighted by difficulty) is mapped onto a bell curve built from a representative norming sample. Scoring at the sample average gives you 100; one standard deviation above gives you 115.

How accurate are IQ tests?

Well-constructed tests are highly reliable, but no score is exact. The standard error of measurement is around 5 points on major tests, so a measured 110 is best read as a range of roughly 105–115. Online tests add further uncertainty because conditions aren't controlled.

Are online IQ tests real?

Good online tests use the same question formats and statistical logic as clinical instruments and can give a useful estimate. But only a full clinical administration — such as the WAIS given one-on-one by a psychologist — produces a score valid for diagnostic or legal purposes.

Can you study for an IQ test?

You can't study the answers, but familiarity helps. Practice with matrix puzzles and timed formats removes the novelty penalty and can lift a score by a few points. Sleep and test conditions matter too. Large, lasting gains from practice alone are not supported by research.

How IQ Tests Work: Scoring & Question Types

IQ tests work by giving you a standardized set of reasoning problems and comparing your performance to a large, representative sample of people your age. You never receive a score of "42 out of 60" — instead, your raw performance is converted into a position on a bell curve where the average is fixed at 100 and the standard deviation is 15. An IQ score is not a measurement of your brain in isolation; it is a statement about where you stand relative to everyone else.

What's actually on an IQ test

Modern tests don't rely on a single question type. They sample several distinct abilities, because intelligence research consistently finds that performance across very different tasks is correlated — the statistical core known as the general factor, or g. The main question families are:

Matrix reasoning

You see a grid of abstract shapes with one cell missing and must pick the option that completes the pattern. This is the Raven's Progressive Matrices format, and it targets fluid reasoning — solving novel problems without relying on learned knowledge. It is the backbone of most serious modern tests, including ours.

Verbal reasoning

Vocabulary, analogies ("hand is to glove as foot is to…"), similarities and comprehension questions. These tap crystallized intelligence — accumulated knowledge and the ability to reason with language.

Working memory

Holding and manipulating information in your head: repeating digit sequences backwards, mental arithmetic, remembering positions in a sequence. Working memory is one of the strongest single correlates of overall reasoning ability.

Processing speed

Simple tasks done under tight time pressure — matching symbols, scanning for targets. Individually trivial, but the speed and accuracy with which you complete them is surprisingly informative about overall cognitive efficiency.

Why matrix reasoning is considered culture-reduced

Verbal questions have an obvious weakness: they depend on the language you grew up speaking and the education you received. A brilliant reasoner tested in their second language will underperform. Matrix puzzles sidestep much of this — they use no words, no arithmetic notation and no culturally specific knowledge. That's why researchers call them culture-reduced (not culture-free — test familiarity and schooling still help), and why matrix formats are preferred when testing across countries or online.

From raw score to IQ: norming and deviation scoring

Here is the machinery that turns puzzle answers into an IQ:

Step 1 — the norming sample.Before a test is published, it is administered to a large sample chosen to represent the population by age, education and region. This sample defines what "normal" performance looks like.
Step 2 — your raw score. When you take the test, your answers produce a raw score — typically the number correct, sometimes weighted by item difficulty.
Step 3 — your position in the distribution.Your raw score is located within the norming sample's distribution for your age group. Scoring exactly at the median puts you at the 50th percentile.
Step 4 — conversion to IQ. That percentile is mapped onto a normal curve with mean 100 and standard deviation 15. The 84th percentile becomes 115; the 98th becomes 130.

This is called deviation scoring, and it is why the average IQ is always 100 by construction, why about 68% of people score between 85 and 115, and why every score has a direct percentile equivalent. The original "quotient" — mental age divided by chronological age — was abandoned decades ago; only the name survives.

Two consequences of this design are worth spelling out. First, an IQ score only means something relative to its norming sample: a score from a test normed in 1985 is not directly comparable to one normed last year, because raw performance drifted upward across the 20th century (the Flynn effect) and publishers must periodically re-norm. Second, the quality of the norms is the quality of the test. A beautifully designed puzzle set normed on a small, self-selected sample — the situation with many free online quizzes — produces numbers that look like IQ scores but aren't anchored to anything.

Your score is a range, not a point

No test measures perfectly. If you took the same test twice, your scores would differ slightly — attention drifts, some items happen to suit you, some don't. Psychometricians quantify this with the standard error of measurement (SEM), which is about 5 points on major tests, sometimes a bit less. In practice, a measured score of 110 means your true score most likely falls somewhere around 105–115. This is why clinicians report confidence intervals rather than bare numbers, and why arguing over a 3-point difference between two people is statistically meaningless.

Retesting adds its own wrinkle: taking the same or a similar test again within months typically raises the score a few points purely through familiarity — a practice effect, not a genuine gain in ability. Serious assessments handle this with alternate forms and minimum retest intervals; serious readers handle it by treating any single sitting as one noisy sample of a stable underlying level.

Reliability vs validity

Two different questions get conflated when people ask whether IQ tests "work":

Reliability asks: does the test give consistent results? Major IQ tests score very well here, with test-retest reliabilities around 0.9 — among the most reliable measurements in psychology.
Validityasks: does the test measure something that matters? IQ scores predict educational attainment, job performance and learning speed better than any other single psychological measure — the APA's task force report confirmed this — but the prediction is probabilistic, not destiny. Plenty of individual outcomes diverge from what the score would suggest.

What IQ tests do not measure

An honest account has to include the boundaries. IQ tests do not measure creativity (generating original ideas is a different skill from converging on one correct answer), emotional intelligence (reading and managing emotions in yourself and others), practical judgment, motivation, curiosity or character. A high scorer can make terrible decisions; an average scorer can build an exceptional career. If you want to raise the abilities the tests do capture, see what the evidence supports in how to increase your IQ.

Online estimates vs clinical administration

A clinical IQ assessment — typically the Wechsler Adult Intelligence Scale (WAIS) — is administered one-on-one by a licensed psychologist over 60–90 minutes, covers ten or more subtests, and produces a formally valid score usable for diagnosis, schooling decisions or legal purposes. It also costs several hundred dollars.

Online tests, including ours, are estimates. A well-built one uses the same matrix-reasoning logic and sound norms, and will land close to a clinical result for most people — but it cannot control your environment, verify your effort or diagnose anything. Treat any online score as a calibrated starting point with a wider error band, not a clinical fact. IQ tests are also just one member of a broader family of assessments used in hiring and education; see our guide to cognitive ability tests for how the pieces fit together, and the IQ scale for what each score range actually means.

How IQ Tests Work: Questions, Scoring and What They Measure