An IQ of 130 means you scored higher than about 97.7% of people your age: in theory, 1 person in 44 scores 130 or more. But that figure assumes a scale with a standard deviation of 15, and every IQ score comes with a margin of error of several points. Here is how an IQ score is built, how to turn it into a percentile and a rarity, and why it should always be read as a range.
- The scale: on the Wechsler scales (WAIS, WISC), the mean is 100 and the standard deviation is 15. The score places you among people of your age.
- Rarity: 115 is about 1 person in 6, 130 about 1 in 44, 145 about 1 in 740.
- Margin of error: for the full-scale IQ, allow 4 to 6 points either side at 95% confidence. A 128 and a 132 are not really different.
- The 130 cut-off: a convention for talking about giftedness, not a diagnosis.
How an IQ score is built
In the early 20th century, IQ was a ratio: mental age divided by real age, multiplied by 100. Lewis Terman used this formula in his 1916 Stanford-Binet. It worked poorly for adults: what "adult mental age" should you divide by?
With his 1939 scale, David Wechsler changed the method. He compared each person with people of the same age. In his own words, an IQ "merely defines his relative position among the group with which he has been compared", and the mean of 100 is simply a historical convention. This is the deviation IQ, still used by the WAIS (adults) and the WISC (children).
In practice, the test measures how many standard deviations you sit above or below the average for your age group (your z-score), then converts it: IQ = 100 + 15 × z. A z-score of +2 gives 130, a z-score of -1 gives 85. If scores follow a bell curve (the normal distribution), about 68% of people score between 85 and 115, and 95% between 70 and 130.
A percentile rank turns the score into the share of people you match or beat: at the 98th percentile, you did as well as or better than about 98% of people your age in the reference sample. It is not a percentage of correct answers. To go from an IQ to a percentile, or back, our IQ and percentile converter does the maths in your browser and stores nothing.
Table: IQ, percentile and "1 person in N"
We computed these values from the normal distribution (mean 100, standard deviation 15). Above 100, rarity means how many people, in theory, score this high or higher. Below 100, this low or lower.
| IQ | z-score | Percentile rank | Share at this score or further from the mean | Theoretical rarity |
|---|---|---|---|---|
| 70 | -2 | 2.3 | 2.3% | 1 in 44 |
| 85 | -1 | 15.9 | 15.9% | 1 in 6 |
| 100 | 0 | 50 | 50% on each side | 1 in 2 |
| 115 | +1 | 84.1 | 15.9% | 1 in 6 |
| 120 | +1.33 | 90.9 | 9.1% | 1 in 11 |
| 130 | +2 | 97.7 | 2.3% | 1 in 44 |
| 145 | +3 | 99.87 | 0.13% | 1 in 741 |
| 160 | +4 | 99.997 | 0.003% | about 1 in 32,000 |
Score reports round these numbers: in a sample report published by the WISC-V publisher, a full-scale IQ of 132 is shown at the 98th percentile. Focus on the order of magnitude. Beyond 145, these figures become extrapolations, as explained below.
Why the same number can mean different rarities
Not every test uses the same standard deviation. The Wechsler scales use 15, as does the Stanford-Binet since its fifth edition (2003). Earlier Stanford-Binet versions used 16, from the switch to the deviation IQ in 1960. And according to a psychologist writing for American Mensa, the Cattell III B test uses 24.
So a "130" does not mean the same thing everywhere (our calculations):
- Standard deviation 15: 130 is 2 standard deviations above the mean, the 97.7th percentile, 1 person in 44.
- Standard deviation 16: 130 is 1.875 standard deviations above, the 97th percentile, 1 person in 33.
- Standard deviation 24: 130 is only 1.25 standard deviations above, the 89.4th percentile, about 1 person in 9.
The reverse is also true: the same rarity gives different numbers. To target the top 2%, Mensa accepts 130 on the Wechsler scales, 132 on older Stanford-Binet versions and 148 on the Cattell. An IQ quoted without its standard deviation cannot be compared with another.
What the manuals call each range
Test manuals attach a label to each score range, and the labels have changed over editions. For the WISC-V (2014), the publisher Pearson uses a symmetrical scale from "Extremely Low" to "Extremely High". The older terms Borderline, Superior and Very Superior, still used by the WAIS-IV, were dropped.
| IQ | WISC-V (2014) | Older terms (WAIS-IV, 2008) |
|---|---|---|
| 130 and above | Extremely High | Very Superior |
| 120 to 129 | Very High | Superior |
| 110 to 119 | High Average | High Average |
| 90 to 109 | Average | Average |
| 80 to 89 | Low Average | Low Average |
| 70 to 79 | Very Low | Borderline |
| 69 and below | Extremely Low | Extremely Low |
Other language editions use their own wording, so go by the words in your own report. The WAIS-5 (2024) uses yet other terms, such as "Above average" for a 111, and Pearson states that these descriptors are only suggestions and are not evidence-based. The borders between labels are conventions of vocabulary, not natural limits. The same is true of any cut-off score, as we explained for the AQ-10 autism questionnaire.
What about low scores, around 70?
An IQ of 70 is as rare as an IQ of 130: about 1 person in 44 scores 70 or lower. According to the American Psychiatric Association, a full-scale IQ of around 70 to 75 indicates a significant limitation in intellectual functioning. But a specific IQ score is no longer required to diagnose intellectual disability: there must also be difficulties in everyday life (adaptive functioning) that began during development, assessed by professionals. A number alone is never enough.
Margin of error: why 128 and 132 are the same
No test measures without error. Tiredness, focus, items that happen to be familiar: the score you get on a given day is only an estimate of your "true" score. This uncertainty is summed up by the standard error of measurement, calculated from the test's reliability: standard deviation × √(1 - reliability). With 95% confidence, the true score lies within 1.96 standard errors of the score obtained.
The Wechsler full-scale IQ is highly reliable: coefficients of .97 to .98 across ages for the WAIS-IV, and .96 to .97 for the WISC-V, according to independent reviews of their manuals. That gives a standard error of about 2 to 3 points, so a 95% confidence interval of about 4 to 6 points either side of the score.
With a standard error of 3 points, a 128 means a range of about 122 to 134, and a 132 a range of about 126 to 138. The two overlap heavily. With this standard error, two scores would need to be about 8 points apart to differ clearly in statistical terms (our calculations).
In the WISC-V sample report published by Pearson, a full-scale IQ of 132 comes with a 95% confidence interval of 125 to 136. It is not centred on 132 because it is built around an "estimated true score", pulled slightly towards the mean. These margins are also a best case: a reviewer of the WAIS-IV notes that they leave out other sources of error, such as changes over the long term and errors in giving or scoring the test. Practice effects add to this: retested a few weeks later, people scored higher on average on the WAIS-IV. France's national health research institute, Inserm, notes that presenting results as confidence intervals can be good practice.
Norms get old: the Flynn effect
An IQ compares you with a reference sample tested at a given time. And average scores rose from one generation to the next for decades. In 1984, James Flynn showed that Americans had gained 13.8 IQ points between 1932 and 1978, about 0.3 points a year. A 2014 meta-analysis (285 studies, 14,031 participants) found an average of 2.3 points per decade, and 2.9 points for the recent versions (since 1972) of the Wechsler scales and the Stanford-Binet.
The practical consequence, according to its authors: a test overestimates scores by about 0.3 points for each year since it was normed. Over 15 years, that would be 4 to 5 points, enough to move someone across the 130 line. This pace is no longer certain: gains have slowed or reversed in several countries, and a 2021 study concluded there is no evidence of a decline in France. Publishers therefore renorm their tests regularly: US norms for the WISC-V were collected in 2013 and 2014, and the WAIS-5 succeeded the WAIS-IV in 2024.
The extremes raise another problem. The US WAIS-IV and WISC-V norm samples each include 2,200 people. In a sample that size, the normal curve predicts about 50 scores above 130, about 3 above 145, and most likely none at 160 (our calculations). Very high scores therefore rest mostly on the statistical model. The full-scale IQ of both tests stops at 160, and reviewers of the WISC-V note that this ceiling may not be enough to identify exceptionally gifted children.
Giftedness and the 130 cut-off: a convention, not a diagnosis
In France, where our site is based, "high intellectual potential" (HPI) is often linked to an IQ of about 130. Inserm describes a score of about 130 as being "considered" high potential. Onisep, the French public agency for school guidance, notes that gifted pupils were long identified on this criterion alone, and that identification now takes psychometric, psychological, educational and medical aspects into account.
There is no real international consensus either, as the French neuropsychologist Laurence Vaivre-Douret points out in a 2025 article. She writes that the threshold, generally a full-scale IQ of at least 130, can be lowered when a neurodevelopmental condition produces an uneven profile of index scores. As for the claim that the World Health Organization set this threshold, Inserm removed it from its page in December 2025 for lack of an official source.
The margin of error also makes a strict line hard to defend: a 127 and a 133 are compatible with the same underlying level. What counts is the full profile, interpreted by a psychologist. A single full-scale IQ can hide quite different strengths, as we explain in Two Kinds of Smart.
What an online IQ test is worth, ours included
A real IQ assessment is done face to face with a psychologist, using a test normed on a representative sample. Online, nobody controls the testing conditions, and the comparison group is made up of the people who chose to take the test.
Our tests are no exception. The complete IQ test (80 questions across four domains) and our IQ tests by domain compare your score with other people who took them on the site, then place it on the usual scale (mean 100, standard deviation 15). The result is an educational estimate, not a clinical IQ, and it cannot tell you whether you are gifted. Our giftedness questionnaire looks at traits often linked with high potential: it is a self-assessment questionnaire, not an IQ test. None of these results is a diagnosis.
Whatever score you have, the IQ and percentile converter gives its percentile, its rarity and a confidence interval, with a standard deviation of 15, 16 or 24. It does not turn an estimate into a clinical measurement, but it helps you see a score as a range.
Frequently asked questions
An IQ of 130 is 1 person in how many?
On a scale with a standard deviation of 15, about 1 in 44: 2.3% of the population scores 130 or more. On a scale with a standard deviation of 24, 130 is only about 1 person in 9.
Is a percentile the percentage of correct answers?
No. A percentile of 84 means you did as well as or better than about 84% of people your age in the reference sample, whatever your number of correct answers.
Why did my IQ change between two assessments?
Several things add up: the margin of error, practice effects if the tests were close together, different tests or norms (the Flynn effect) and, in children and teenagers, a real change compared with others of the same age. A difference of a few points is not surprising.
Does a score of 135 on an online test mean I am gifted?
No. An online test gives an estimate, without supervision or a representative sample. Only an assessment by a psychologist can speak to giftedness, and even then, 130 remains a convention.
References: Wechsler D (1944). The measurement of adult intelligence, 3rd edition (1st edition 1939). Baltimore: Williams & Wilkins, chapter 3, pp. 34-35 | Roid GH, Barram RA (2004). Essentials of Stanford-Binet Intelligence Scales (SB5) Assessment, chapter 1. Hoboken: Wiley | Salny AF. The question of "genius". American Mensa (accessed 10 November 2026) | American Mensa. Qualifying test scores and Join American Mensa (top 2% requirement, accessed 10 November 2026) | Pearson (2020). WISC-V Interpretive Report, sample report | Pearson (2008). WAIS-IV Score Report, sample report | Pearson (2024). WAIS-5 Score Report, sample report and WAIS-5 product page | Pearson (2019). WISC-V UK brochure, table of descriptive classifications (traditional and WISC-V terms) | Canivez GL (2010). Review of the Wechsler Adult Intelligence Scale, Fourth Edition. The Eighteenth Mental Measurements Yearbook. Lincoln, Nebraska: Buros Institute, pp. 684-688 (independent review of the technical manual; norm sample of 2,200 people) | Canivez GL, Watkins MW (2016). Review of the Wechsler Intelligence Scale for Children, Fifth Edition: critique, commentary, and independent analyses. In: Kaufman AS, Raiford SE, Coalson DL. Intelligent testing with the WISC-V. Hoboken: Wiley, pp. 683-702. doi:10.1002/9781394259397.ch20 (independent review; norm sample of 2,200 children, collected April 2013 to March 2014) | Harvill LM (1991). Standard error of measurement. Educational Measurement: Issues and Practice 10(2):33-41. doi:10.1111/j.1745-3992.1991.tb00195.x | American Psychiatric Association. What is intellectual disability? (reviewed March 2024) | Flynn JR (1984). The mean IQ of Americans: massive gains 1932 to 1978. Psychological Bulletin 95(1):29-51. doi:10.1037/0033-2909.95.1.29 | Trahan LH, Stuebing KK, Fletcher JM, Hiscock M (2014). The Flynn effect: a meta-analysis. Psychological Bulletin 140(5):1332-1360. doi:10.1037/a0037173 (meta-analysis, 285 studies, 14,031 participants) | Bratsberg B, Rogeberg O (2018). Flynn effect and its reversal are both environmentally caused. PNAS 115(26):6674-6678. doi:10.1073/pnas.1718793115 (Norwegian conscription data, cohorts born 1962 to 1991) | Gonthier C, Grégoire J, Besançon M (2021). No negative Flynn effect in France: why variations of intelligence should not be assessed using tests based on cultural knowledge. Intelligence 84:101512. doi:10.1016/j.intell.2020.101512 | Ramsden S, Richardson FM, Josse G, et al. (2011). Verbal and non-verbal intelligence changes in the teenage brain. Nature 479(7371):113-116. doi:10.1038/nature10514 (longitudinal study, 33 adolescents tested at ages 12 to 16 in 2004, then in 2008) | Inserm (10 November 2022, corrected 2 December 2025). Identifier le haut potentiel intellectuel avec un test sur Internet, vraiment ? (in French) | Onisep. Élèves à haut potentiel : une aide au repérage (in French, accessed 10 November 2026) | Vaivre-Douret L (2025). Identification du haut potentiel intellectuel, conduite à tenir. Journal de pédiatrie et de puériculture 38(2):59-71. doi:10.1016/j.jpp.2024.10.003 (in French)