An IQ score is not the percentage of questions you answered correctly. Modern scoring converts your raw performance into a standard score on a normal distribution with a mean of 100 and a standard deviation of 15, by comparing your results to a representative norm sample of people your age. The number reflects your rank relative to others, not raw points.
Key takeaway: IQ scoring ranks you against a norm group rather than counting correct answers. Your raw score is mapped onto a bell curve (mean 100, SD 15), often using item-response theory so that harder questions count for more. An honest result reports a confidence range and a percentile, not a single exact number, and an online estimate is never a clinical or Mensa diagnosis.
How is IQ calculated?
The intuitive guess is that IQ equals "percent correct," but that is not how any serious test works. If a test simply counted right answers, the same raw score would mean wildly different things depending on how hard the questions were, how old you are, and who else took the test. Scoring solves this by translating your raw performance into a position on a shared scale.
The process generally has three stages:
- Collect a raw score. You answer the items, and the test records your pattern of responses — which ones you got right, which you missed, and often how hard each one was.
- Compare you to a norm sample. Before the test was published, it was given to a large, carefully chosen group of people. Your performance is measured against people in your age band, because cognitive performance changes across the lifespan.
- Convert to a standard score. Your relative standing is mapped onto the normal distribution, producing a number centered on 100. This is the figure people call "IQ."
IQ is a relative measurement. A score of 100 means you performed about average for your age group. It does not mean you answered exactly 100 questions, or 100% of them. The number describes where you sit compared to everyone else, not how many points you piled up.
This is why the same person can take two tests, get a different number of items correct, and still land near the same IQ. The norm group and the scale, not the raw count, decide the final score. You can see honest scoring in action on our free IQ test, which reports an estimate with a confidence range and a percentile rather than an inflated single number.
What are the bell curve and standard scores?
The "bell curve," or normal distribution, is the statistical shape that describes how many human traits — including measured cognitive performance — spread across a population. Most people cluster near the middle, and fewer people appear as you move toward the high and low extremes. IQ scales are deliberately built to fit this curve.
By convention, the modern IQ scale is anchored to two numbers:
- Mean of 100. This is the center of the distribution, defined as average performance for an age group.
- Standard deviation of 15. The standard deviation measures how spread out scores are. One standard deviation above the mean is 115; one below is 85.
Because the curve is fixed, every IQ number corresponds to a predictable slice of the population. Roughly two-thirds of people fall between 85 and 115 — within one standard deviation of the mean. The further a score sits from 100, the rarer it becomes, which is exactly why very high and very low scores are uncommon by design.
A "standard score" simply re-expresses your raw result in units of standard deviations from the mean. That is what lets a single number on the bell curve carry meaning: it tells you not just that you are above or below average, but roughly how far and how rare that standing is. If you want to see how the curve maps onto specific bands, our guide to IQ score ranges breaks them down.
What is item-response theory?
Older scoring approaches treated every question as equally informative: one right answer, one point. Most well-built modern tests use something more sophisticated called item-response theory, usually shortened to IRT. Instead of just tallying correct answers, IRT models the relationship between your underlying ability and the difficulty of each individual item.
Two ideas sit at the heart of IRT:
- Harder items carry more weight. Correctly answering a difficult question is stronger evidence of high ability than answering an easy one, so it influences your estimate more.
- The test estimates the ability that best fits your whole pattern. Rather than reading a single total, the scoring engine looks at your entire set of responses and asks: what level of ability would most likely produce this pattern of right and wrong answers? That best-fitting estimate becomes the basis of your score.
Imagine two people each answer ten questions correctly. One got the ten easiest items right and missed the hard ones; the other solved several of the hardest items but slipped on some easy ones. A raw count says they tied. IRT does not: it reads the difficulty pattern and can estimate a higher ability for the person who cleared the harder problems.
A useful side effect of IRT is that it naturally produces a measure of uncertainty. Because the model is estimating an unseen ability from limited evidence, it can also report how precise that estimate is — which leads directly to the idea of a confidence range.
Why a confidence range instead of one number?
No test measures ability perfectly. Your performance on any given day is nudged by fatigue, attention, mood, guessing, test conditions, and plain luck on which specific questions appear. Because of this, the honest output of a test is not a single exact figure but a range in which your true ability most likely sits.
The technical name for this is the standard error of measurement. It quantifies how much your observed score might wobble if you took an equivalent test again under similar conditions. A responsible report uses it to present something like "your estimate is around 112, most likely between 105 and 119" rather than declaring a flat, falsely precise "112."
Treat any test that hands you a single dramatic number with zero margin of error with suspicion — especially free online tests promising sky-high results. A score with no confidence range hides the uncertainty that genuinely exists. We dig into this further in whether online IQ tests are accurate.
Reporting a range is not a weakness of the test; it is a sign of honesty. It tells you the measurement was taken seriously and that the result should be read as an estimate of where you stand, not as a fixed, permanent label.
How does honest scoring work?
Putting the pieces together, honest scoring is less about producing an impressive headline number and more about giving you an accurate, well-framed picture of your standing. A trustworthy result tends to share a few features.
- It compares you to a norm sample, not to an absolute scale. Your score reflects your rank within a relevant population, adjusted for age.
- It uses the bell curve transparently. The mean-100, SD-15 framework lets the number map cleanly onto a percentile, so you can see what proportion of people you scored above.
- It applies item-response theory where possible. Difficulty-weighted scoring extracts more information from your answers than a flat tally ever could.
- It reports a confidence range. The standard error is shown, so you understand the precision of your estimate.
- It stays modest about what it is. An online estimate is a useful, fun snapshot — not a clinical assessment and not a Mensa qualification, both of which require supervised, professionally administered tests.
The most honest thing a score can tell you is your percentile — the share of the norm group you outperformed — together with the range around it. A percentile is harder to inflate and easier to interpret than a bare number. Our explainer on IQ percentiles walks through how to read one.
Read this way, an IQ result becomes a reasonable estimate of relative cognitive performance on a particular kind of task, on a particular day — informative, but never the whole story of a person's mind.
How is IQ measured?
IQ is measured by giving you a set of standardized cognitive tasks, recording your pattern of responses, and comparing that performance to a norm sample of people in your age group. Your relative standing is then converted into a standard score on the bell curve, centered on 100 with a standard deviation of 15. It is a measure of rank, not a count of correct answers.
What is a standard deviation in IQ?
A standard deviation describes how spread out scores are around the average. On the standard IQ scale it equals 15 points. So a score of 115 is exactly one standard deviation above the mean of 100, and 85 is one below it. About two-thirds of people score within one standard deviation of average, which is why most results land between 85 and 115.
Why do I get different scores on different tests?
Different tests use different norm samples, different question types, and sometimes different scoring methods, so the same person can land on different numbers. Day-to-day factors like fatigue, attention, and guessing also shift results, which is captured by the standard error of measurement. This is exactly why an honest report gives a confidence range, and why an online score should be read as an estimate rather than a definitive figure.