In short
The IQ scale sets the average of a reference group at 100 and makes one standard deviation 15 points. That is a construction, not a discovery: scores are ranked, then stretched onto a normal curve, which is why about 68 percent of people fall between 85 and 115 and about 95 percent between 70 and 130, and the familiar range labels are publishing conventions that differ between instruments and have no natural cut-points behind them. A professional report never gives a bare number: it gives a percentile and a confidence interval, because the same person retested will land somewhere nearby rather than on the same figure. Psychology.com's IQ Test reports answers correct out of 30 with an estimated performance range, not an IQ score, and it is a reasoning quiz rather than a clinical assessment.
- Average
- 100, by construction rather than by discovery
- One step
- 15 points is one standard deviation
- Middle of the curve
- About 68 percent of people fall between 85 and 115
- Range labels
- Publisher conventions, not natural categories
- Always read with
- A percentile and a confidence interval
- Test
- IQ Test: Free Reasoning Quiz
The scale was built, not found
The first intelligence quotients really were quotients. A child's measured mental age was divided by their actual age and multiplied by a hundred, so a nine-year-old performing like an average eleven-year-old scored about 122. That arithmetic breaks down in adulthood, because mental age stops climbing while chronological age does not, and it produced scores that were not comparable from one age to another.
David Wechsler replaced it in 1939 with the deviation approach every modern test uses. Instead of dividing ages, you take a large sample of people in a given age band, rank everybody's raw performance, and convert the ranks onto a scale whose average is set to 100 and whose standard deviation is set to 15. The number you receive says where you sit relative to that sample. It is not a count of anything.
Two consequences follow immediately. First, the average is 100 because somebody decided it would be, and it will be 100 again in the next edition no matter how the raw performance of the population moves. Second, every score belongs to a particular reference sample, collected in a particular country in a particular decade. A score without that context is a number without a meaning.
What the bell curve actually says
Because scores are mapped onto a normal distribution, how common any score is follows from the mathematics rather than from any finding about people. The figures below are our own calculation from a normal curve with a mean of 100 and a standard deviation of 15. They are not taken from any publisher's conversion table, and the real reference data for a given test departs slightly from the ideal curve at the extremes.
- A score of 70 sits at roughly the 2nd percentile: about 2 in 100 people in the reference sample scored at or below it.
- A score of 80 sits at roughly the 9th percentile, and 85 at roughly the 16th.
- A score of 90 sits at roughly the 25th percentile, which is to say a quarter of the sample scored lower.
- A score of 100 is the 50th percentile by definition: half above, half below.
- A score of 110 sits at roughly the 75th percentile, and 115 at roughly the 84th.
- A score of 120 sits at roughly the 91st percentile, and 130 at roughly the 98th.
- A score of 145 sits at roughly the 99.9th percentile, about 1 person in 740.
- Between 85 and 115 lies about 68 percent of the sample; between 70 and 130, about 95 percent.
The middle of the scale is crowded and the edges are empty, which is the most useful fact on this page. Ten points near 100 move you across a large slice of the population, while ten points near 140 move you past almost nobody, because there is almost nobody left out there to move past. This is why differences near the average mean less than they look like they mean, and why tests lose precision at the top and bottom.
The range labels are conventions
Score reports attach words to bands: average, above average, below average, and so on at the extremes. Those words are editorial decisions made by a publisher, and they are not the same from one instrument to the next. A score two points either side of a boundary earns a different word while describing performance that is, for practical purposes, identical. Publishers know this, which is why the manuals describe the labels as descriptive classifications rather than categories.
There is nothing natural about the cut-points either. The distribution is a smooth curve with no gaps in it, so any line drawn across it is drawn by a person. Dichotomizing a continuous score into a band discards information and can create apparent differences between people who are a rounding error apart, a general measurement problem rather than something peculiar to intelligence testing.
The honest way to use a label is as shorthand for a percentile you have already looked at. If the report says a score sits at the 37th percentile and calls it average, the percentile is the fact and the word is the summary. If you only have the word, you are working with the least informative part of the result.
A score is a band, not a point
No test measures without error, so every score has a margin. A professional report states that margin around each index and around the total: every composite comes with a percentile rank, a 95 percent confidence interval and a standard error of measurement, and those intervals typically span about ten points.
That interval is not a disclaimer. It is the result. If your full-scale score is 94 with an interval of 89 to 99, the finding is that your true standing is somewhere in that span, most plausibly near the middle. Comparing your 94 with a friend's 99 is comparing two bands that mostly overlap, and a five-point gap is no bigger than the margin on either score.
The same logic governs retesting. Repeat the test a month later and the number will move a few points in one direction or the other, partly because people vary and partly because measurement does. Practice is a real effect too: a meta-analysis of retesting on cognitive ability tests found a consistent gain from simply having sat the test before, larger when the identical form is reused. Any single figure should be held loosely.
What a high score does and does not imply
A score near the top of the scale means the person handled an unusually large share of the test's items correctly relative to their age group. That is informative, and in some settings it is useful: it can support an application for advanced coursework, and at a group level higher scores are associated with more years of education and with occupations that require more training.
What it does not do is forecast an individual life. The correlations behind the group-level claims leave an enormous amount unexplained, and a longitudinal meta-analysis found that parental background and school grades predicted later educational attainment almost as strongly as measured ability did. Two people fifteen points apart routinely finish in the reverse order on anything you would actually want to know about.
Tests also lose their grip at the top. Reference samples contain relatively few people above about 130, so the scale up there rests on thin data, and most batteries do not have enough hard items to separate one very high scorer from another. That is why our own quiz stops short of awarding exceptional-ability labels near its ceiling: the instrument has run out of information before the person has.
What a low score does and does not imply
A low score on one occasion establishes very little by itself. Unfamiliar language, limited experience with written arithmetic, poor sleep, pain, medication, anxiety, a noisy room or a misunderstood instruction will all pull a result down without telling you anything about the person's capability on a better day. This is one of the strongest reasons that a bare number is never a conclusion.
Where a formal determination is involved, the criteria are deliberately wider than a score. The American Association on Intellectual and Developmental Disabilities describes three requirements together: significant limitation in intellectual functioning, significant limitation in adaptive behavior across conceptual, social and practical skills, and onset during the developmental period before age 22, with a clinician reaching a determination only after a comprehensive evaluation. A test result is one input into that work and is not a diagnosis on its own.
If a low result on any test has unsettled you, the useful response is not to retake it until the number improves. It is to describe the actual difficulty, the thing you noticed in your work or your day, to someone who can assess it properly. A change from your own usual functioning matters far more than where a single score fell on a curve.
Why online scores run high
Free online tests hand out generous numbers, and the reasons are structural. Many have no reference sample at all, so whatever scale they print is invented. Others calibrate against their own visitors, who are self-selected people interested in reasoning puzzles rather than a slice of the public, which shifts the apparent average upward and makes an ordinary performance look exceptional.
Commercial incentives push the same way. A result that flatters gets shared, and a site selling a certificate or a report has no reason to tell you that you landed near the middle. Untimed, unproctored conditions add their own inflation: breaks, notes, second attempts and search engines are all available, and none of them are available in a clinical session.
The test for whether an online number means anything is simple. Ask what sample it was compared against, when that sample was collected, and who collected it. If the page cannot answer, the number is a score on that page's questions and nothing more. Ours is in that category, and we say so rather than implying otherwise.
How to read our own result
Psychology.com's IQ Test gives 30 original text-only questions, untimed, with no speed bonus, and reports an estimated range plus a breakdown across five categories of reasoning. We deliberately report a range rather than a point, and we do not convert it into an IQ figure, because the quiz is self-administered, unproctored and unnormed. It is not an IQ score and not a clinical assessment.
The breakdown is the part worth reading. The categories hold different numbers of items and have not been matched for difficulty, so they are a description of how this attempt went rather than a profile of your abilities. Still, if logical items went smoothly and numerical items did not, that is a reasonable prompt to practice one kind of reasoning rather than a verdict about either. An optional in-depth report is available if you want a longer write-up of the result.
If what you actually want is a percentile against other adults, no free online test can supply it, ours included. That comparison requires a representative reference sample and a qualified examiner, and the only route to it is a professional appointment.
What the research says
The measurement side of this is solid. Modern batteries are among the most reliable instruments in psychology, the normal-curve scaling is a well-understood piece of statistics, and the group-level associations between scores and educational, occupational and health outcomes have been replicated in large prospective samples. None of that is in serious dispute.
What the scale cannot do is treat its own numbers as fixed facts about people. Average raw performance on the same tests rose through most of the twentieth century, by roughly three points a decade in a meta-analysis pooling 271 samples from 31 countries, which is why every edition is re-normed and why a score from 1985 and a score from 2025 are not the same currency. In several countries that rise has stalled or reversed.
So the scale is best understood as a well-made ruler for a quantity that keeps moving, read against a reference group that keeps being replaced. Used that way it is genuinely informative. Used as a permanent label, it claims a precision that no part of the machinery behind it supports.
| Feature | What the number means | What it is often taken to mean |
|---|---|---|
| A score of 100 | The midpoint of the reference sample | Being unremarkable |
| A score of 130 | Roughly the 98th percentile of that sample | A certified level of brilliance |
| A score of 85 | Roughly the 16th percentile, a common result | Evidence of a problem |
| A five-point gap | Inside the margin of error on most reports | A meaningful difference between two people |
| A range label | A publisher's word for a band of scores | A natural category people belong to |
| A score from 1990 | A rank against a 1990 reference sample | Directly comparable with a score today |
| An online result | Performance on that site's questions | A percentile among adults |
Key takeaways
- The scale is a construction: the reference group's average is set to 100 and one standard deviation is set to 15, so the average is 100 by definition rather than by finding.
- On a normal curve with those settings, about 68 percent of people fall between 85 and 115 and about 95 percent between 70 and 130, which is why differences near the middle mean less than they look like they mean.
- The familiar range labels are publisher conventions with no natural cut-points behind them, and they differ between instruments.
- Every score carries a margin. Read the percentile and the 95 percent confidence interval, and treat a gap of a few points between two scores as no gap at all.
- Online scores run high because many have no reference sample, calibrate against self-selected visitors, or are taken under conditions nobody controlled.
See how you do on the reasoning itself
Our free IQ Test gives 30 untimed text questions and reports an estimated range with a five-category breakdown, deliberately without converting it into an IQ figure.
Cite this source
Fontane Pennock, S. (2026, November 8). IQ Scale: Score Ranges and the Bell Curve. Psychology.com. https://psychology.com/types/iq-scale
Frequently asked questions
What is the IQ scale, and why is the average 100?
It is a ranking converted onto a curve. A large reference sample of people in your age band sit the test, everyone's raw performance is ranked, and the ranks are rescaled so the sample's average becomes 100 and one standard deviation becomes 15 points. The average is 100 because the scale is defined that way, and it will be 100 again in the next edition regardless of how raw performance has moved.
What counts as a high IQ score?
On a normal curve with an average of 100 and a standard deviation of 15, a score of 130 sits at roughly the 98th percentile and 145 at roughly the 99.9th. Those are the figures usually behind the words. Treat them as positions in a reference sample rather than thresholds: the test has few hard items up there, so it cannot separate one very high scorer from another with any confidence.
What does an IQ of 120 mean?
On the ideal curve it sits at roughly the 91st percentile, meaning about nine in ten people in the reference sample scored lower. In practice a report would give it with a confidence interval of roughly ten points, so the honest statement is that the person's standing is somewhere between the mid-110s and the mid-120s. It implies nothing about any particular skill, qualification or outcome.
Why is the IQ bell curve shaped like that?
Partly because many small influences on performance add up, which tends to produce a normal distribution, and partly because the scores are explicitly mapped onto one. The ranking step comes first and the curve is imposed second. Real reference data is close to normal in the middle and departs from it at the extremes, which is one reason very high and very low scores are measured least precisely.
Is a 5 or 10 point difference in IQ meaningful?
Five points, no. Ten points, only just. Reports give each score a 95 percent confidence interval spanning roughly ten points, so a five-point gap sits inside the margin of both. Retesting the same person also shifts the number, and a meta-analysis of retesting found a reliable practice gain from simply having sat the test before. On those figures a gap between two people has to reach about seven points before it clears the combined margin of both scores.
Do the IQ range labels mean the same thing on every test?
No. The words attached to bands are editorial choices made by each publisher, and the boundaries differ between instruments and between editions of the same instrument. The underlying distribution has no gaps, so every cut-point is drawn by a person. Use the percentile as the fact and treat the label as shorthand for it.
Why do online IQ tests give such high scores?
Most have no representative reference sample, so any scale they print is invented. Those that calibrate against their own visitors are comparing you with people who chose to look up a reasoning quiz, which shifts the apparent average. Unproctored conditions, repeat attempts and a flattering result that gets shared all push the same way. Ask what sample a number was compared against before trusting it.
Related types
References
- Wechsler D. The Measurement of Adult Intelligence. Williams and Wilkins; 1939. (The deviation approach that replaced the mental-age quotient.) openlibrary.org
- Boake C. From the Binet-Simon to the Wechsler-Bellevue: Tracing the History of Intelligence Testing. Journal of Clinical and Experimental Neuropsychology. 2002;24(3):383-405. doi.org
- Hausknecht JP, Halpert JA, Di Paolo NT, Moriarty Gerrard MO. Retesting in selection: A meta-analysis of coaching and practice effects for tests of cognitive ability. Journal of Applied Psychology. 2007;92(2):373-385. doi.org
- Pietschnig J, Voracek M. One Century of Global IQ Gains: A Formal Meta-Analysis of the Flynn Effect (1909-2013). Perspectives on Psychological Science. 2015;10(3):282-306. doi.org
- Strenze T. Intelligence and socioeconomic success: A meta-analytic review of longitudinal research. Intelligence. 2007;35(5):401-426. doi.org
- American Association on Intellectual and Developmental Disabilities. Defining criteria for intellectual disability. aaidd.org
