The IQ scale and score chart: what 100, 110, 115, 120 and 130 mean

IQ score chart

The chart below applies the usual IQ scale, with a mean of 100 and a standard deviation of 15, to each band of scores. It shows the share of people in each band, and the percentile at each band's lower edge. The figures are the normal curve and nothing else. They describe the scale a test is built on, not any particular group of people, and every real score comes with an error band of several points, so a score near the edge of a band could belong to the next one.

Two landmarks sit inside a band: 110 is about the 75th percentile, with about 1 in 4 scoring higher, and 120 is about the 91st percentile, with about 1 in 11 scoring higher.

Two things the chart does not do. It does not label the bands: manuals use different words for the same range, so the words are left to the test that printed your score, and the section on range labels below explains why. And it does not turn a raw count of correct answers into an IQ. Only a test's own norm sample can do that.

What each score means, from 70 to 145

A single IQ number answers one narrow question: where a performance sat in the group the test was normed on. One standard deviation on the usual scale is 15 points, so the distance between the score and 100 fixes the share of people at or below it, and that share is the percentile. Two things belong with every figure below and are almost always missing from the pages that quote them. A real test report prints the score inside a confidence band of several points rather than as a point, and the number belongs to the test that produced it, because each test is scaled on its own norm sample. A score with no test named beside it is an incomplete sentence.

Below 70: about 2% of the scale

About 2.3% of a norm sample scores below 70, two standard deviations under the middle. A score here is never a diagnosis on its own; the section on 70 below explains what an assessment adds.

70 to 84: about 14% of the scale

About 13.6% of a norm sample scores from 70 up to 85, between one and two standard deviations below the middle.

70: about the 2nd percentile

A score of 70 is two standard deviations below the middle of the scale, which places it at about the 2nd percentile: about one person in 44 in the norm sample scored at or below it. It is the same distance below 100 that 130 sits above it, and it describes the same share of people. It is not a diagnosis of anything. Assessments that use a score in this region combine it with a developmental history, an interview and a measure of how someone manages everyday life, and none of those three is a number.

85 to 99: about 34% of the scale

About 34.1% scores from 85 up to 100, within one standard deviation below the middle. With 100 to 114, this band holds about 68% of everyone.

85: about the 16th percentile

A score of 85 is one standard deviation below the middle, which places it at about the 16th percentile: roughly one person in six scored at or below it. Together with 115 it marks the edge of the band that holds about 68% of a norm sample, which is why the range 85 to 115 is quoted so often. With a confidence band of several points around it, 85 and 89 are similar scores rather than two distinct results.

100 to 114: about 34% of the scale

About 34.1% scores from 100 up to 115, within one standard deviation above the middle, the mirror image of the band below.

100: the middle of the scale

A score of 100 is the 50th percentile, the point half the norm sample scored at or below, and it sits there because the publisher put it there rather than because anybody discovered it. The scores people look up around it land close by: 103 is about the 58th percentile and 108 about the 70th, and both sit inside the band most manuals call average. A gap of a few points in this region is smaller than the error band on either score, so it is not a difference at all.

110: about the 75th percentile

A score of 110 is two thirds of a standard deviation above the middle, which places it at about the 75th percentile: higher than roughly three people in four in the group the test was normed on. Most current manuals put 110 to 119 in a band they call high average, and that is a reporting convention rather than a description of a person. On a second sitting of the same test the same person might print 106 or 114 without anything about them having changed, which is the single most useful thing to know about a number in this range.

115 to 129: about 14% of the scale

About 13.6% scores from 115 up to 130, between one and two standard deviations above the middle; 120, about the 91st percentile, sits inside it.

115: about the 84th percentile

A score of 115 is exactly one standard deviation above the middle, which places it at about the 84th percentile: higher than roughly five people in six. That is where the phrase 84th percentile on so many report forms comes from, and it is arithmetic rather than a finding about anyone. On a test scaled with a standard deviation of 16 the same percentile is 116 instead, which is the smallest clear illustration of why the test has to be named beside the number.

120: about the 91st percentile

A score of 120 is one and a third standard deviations above the middle, which places it at about the 91st percentile: higher than roughly nine people in ten. Manuals disagree about what to call the band it sits in, with superior and well above average both in current use and the wording changing between editions of the same test. The error band matters here as much as anywhere: on the confidence intervals a real report prints, 118 and 122 are not two different results.

130 to 144: about 2% of the scale

About 2.1% scores from 130 up to 145.

130: about the 98th percentile

A score of 130 is two standard deviations above the middle, which places it at about the 98th percentile: higher than about 43 people in 44. This is the region the high-IQ societies draw their line in. Mensa International states that the only qualification for membership is scoring within the upper 2% of the general population on an approved intelligence test, and which number meets that line depends entirely on the test: American Mensa's published list accepts 130 on the Stanford-Binet 5, 132 on the earlier Stanford-Binet and 148 on the Cattell, all for the same 2%.

145 and above: about 1 in 740

About 0.13% scores 145 or more, three standard deviations above the middle, and this is where a printed number is least precise.

145: about the 99.9th percentile

A score of 145 is three standard deviations above the middle, which places it at about the 99.9th percentile: about one person in 740 scores higher. On the way there, 140 is about the 99.6th percentile, roughly one person in 260. Popular labels such as genius attached to 140 or 145 are not part of any current test manual. This is also the region where a printed number is least trustworthy, because few items in a test are hard enough to separate people at this end of the curve, so the confidence band widens exactly where the figure tends to be quoted most precisely.

Every score is a band, not a point

The same person can score 112 on one good test and 119 on another a month later, and both results are the same finding. A professional report does not print a bare number. Pearson's own published sample report for the WAIS-IV gives a Full Scale IQ of 139 alongside a percentile rank of 99.5, a 95% confidence interval of 134 to 142, and a one-word qualitative description, and it prints the standard error of measurement for every scale beside it, running from about two points on the Full Scale to four and a half on Processing Speed. The Institute of Medicine's overview of psychological testing gives the reason in one clause, that obtained test scores are only estimates of true scores.

This is the fact that the pages built around a single number leave out, and it is the one that changes what to do with the number. A score is a most likely value inside a range. If the range matters to a decision, the decision needs the report, the test name, the date and the interval, and not the headline figure on its own.

The same standing is a different number on a different test

A number without its test is incomplete, because every test is scaled on its own norm sample and some are not scaled the same way at all. Most current tests use a standard deviation of 15. Some editions of the Stanford-Binet used 16, which stretches the scale: a score of 132 on an SD-16 test sits at the same percentile as 130 on an SD-15 test. The converter at the top of this page takes either, because the right value is the one printed in the report you are holding.

Editions matter as much as publishers. Measured performance on the same old items has drifted upward over decades in many countries, the pattern usually called the Flynn effect: a review in the Indian Journal of Psychiatry puts it at roughly 10 to 20 points of population IQ every thirty years or so, and calls it a universal finding. Each new edition of a test is re-normed against a fresh sample, which sets its middle back to 100. So a score of 100 on a 1980 edition and a score of 100 on a current edition both mean the middle of the group that took that edition, and not the same number of solved items. American Mensa accepts scores from approximately 200 different standardized intelligence tests and requires that each be administered under conditions appropriate for the norming standards of that test. Each test is its own scale, and a number carried across from one to another is a translation rather than a reading.

Why the range labels differ between tests

Test manuals divide the curve into bands and give them names, and the names are not interchangeable. One publisher calls 90 to 109 average and 110 to 119 high average; another uses superior where a third says well above average; older manuals used labels that are no longer considered acceptable. The bands are conventions for reporting, not categories of person, and the same score can carry a different word depending on which manual printed it. When you compare two reports, compare percentiles rather than labels.

Where these percentiles come from

Every figure on this page is the normal distribution and nothing else. Once the mean is set to 100 and the standard deviation to 15, the landmarks are arithmetic: one standard deviation above the mean is 115, two is 130 and three is 145, and the same distances below give 85, 70 and 55. About 68% of a norm sample falls between 85 and 115, about 95% between 70 and 130, and about 2% above 130. Because the curve is symmetric, the share of people above 115 equals the share below 85.

None of that arithmetic is a claim about people. It describes how a set of raw scores was transformed so that they could be compared with one another, which is what norm-referenced means: such measures inform the test-takers where they stand relative to others in the distribution. The distribution is the thing being described. The person is not.

How to use a score

Our reasoning tests show your answers by skill, with explanations; the IQ-style forms also show a practice estimate. This converter is for a standardized score you already have: it shows the percentile implied by the selected scale.

If a score will inform a decision, bring the full report: the test name and edition, date, percentile and confidence interval. Those details show which norm group the score uses and how much measurement uncertainty to allow.