
Cox rated him 165 for ages 17 to 26, 2 points below 167. Cox's childhood rating for him was 180.
A 167 IQ means a score higher than about 99.9996% of people, roughly 1 in 251,738. It is also the lowest of the figures that have been reported for Marilyn vos Savant, whose childhood scores range from 167 to 230 depending on who did the arithmetic.
At 67 points above the mean of 100 (4.47 standard deviations), 167 sits where the curve has almost flattened onto the axis. Only 0.0004% of people score here or higher; on the left chart that group is the barely visible orange edge.
At 167, the share of people further out is about four in ten million, a sliver too thin to see on a full-size chart.
Vos Savant was listed by Guinness from 1985 to 1989 with an IQ of 228, from a Stanford-Binet test taken at age 10 and scored with the old "mental age divided by real age" formula. As our article on the highest IQ in the world explains, a school record gives a lower mental age from a later sitting, recalculations with different ages give 218, and 228 is often rounded up to 230.
The spread from 167 to 230 is a lesson about method, not about her. Ratio scores like 228 can't be placed on today's bell curve, and a single childhood test can be re-scored in more than one defensible way. A modern deviation score of 167, by contrast, means one thing: a rank of about one in a quarter of a million.
How big a group do you need before someone in it is likely to score 167? About 251,738 people. Watch the crowd grow to that size.
One dot per person, and on average only one of them, shown in orange, scores 167 or more.
| Worldwide (8.2 billion people) | about 32,600 |
| India | about 5,820 |
| United States | about 1,360 |
| United Kingdom | about 276 |
| Canada | about 165 |
| Australia | about 110 |
Population: World Bank, total population, 2025. Counts assume the same bell curve in every country and are rounded. Real tests can't sort people this finely, as the next sections explain.
Entry to a high-IQ society depends on a percentile, which each one turns into a score per accepted test. Measured under supervision, 167 meets 5 of the 6 bars; the next, Mega Society, needs about 171.
Scores for Mensa, Intertel, Triple Nine and ISPE are the Wechsler figures those societies list or that match their percentile. Prometheus and Mega give only a rarity; the scores shown are the matching points on the standard scale. Prometheus says no test currently being given is approved for its entry. Online test results are not accepted by any of them.
The adult Wechsler test and the Stanford-Binet 5 both stop at 160, so neither can give a 167. To report a score this high, a psychologist needs an extended scale, such as the WISC-V extended norms (to 210) or the Stanford-Binet 5 Extended IQ (to 225), shown in the bars below.
Why so uncertain? Because a test compares you with its norm sample, and the WAIS-IV sample had 2,200 adults. The bell curve expects about 0.0087 of them at 167 or higher. With nobody in the sample anywhere near this level, the number comes from the shape of the curve, not from comparison with real people.
There is measurement error too: a professional report would give a range of about 162 to 172 around a 167. With so few people ever measured at this level, the true uncertainty is larger still. Online results are far rougher, which is why ours stops at 145. Are online IQ tests accurate?
Sources: WAIS-IV range and norm sample from Canivez (2010), Mental Measurements Yearbook review; Stanford-Binet 5 and its Extended IQ from PRO-ED; WISC-V extended norms from Pearson Technical Report #6 (2019). Striped bars are extended scales, used only when a person hits the normal ceiling.
The number attached to a result depends on the scale's standard deviation. Near 167, switching scales can add or remove dozens of points for exactly the same rank.
So a "167" isn't always a 167. On the Cattell scale it corresponds to about 142 on the standard scale; on an SD 16 scale, about 163. Why IQ scales differ.
Catharine Cox's 1926 study is the source of nearly every historical "IQ". She estimated each person's score from records of their early life, once for childhood and once for ages 17 to 26. None of them sat a test.

Cox rated him 165 for ages 17 to 26, 2 points below 167. Cox's childhood rating for him was 180.

Cox rated him 165 for ages 17 to 26, 2 points below 167. Cox's childhood rating for him was 175.
Cox rated him 165 for ages 17 to 26, 2 points below 167. Cox's childhood rating for him was 145.
Source: Cox, C. M. (1926), Genetic Studies of Genius, Vol. II: The Early Mental Traits of Three Hundred Geniuses, Stanford University Press. Cox stressed that thin records pull estimates down. Read more in where the historical genius IQ estimates came from.
Under the old ratio formula, IQ equals mental age divided by real age, times 100. The answer depends on which mental age you take from the test record and which age you divide by, and for a ten-year-old a difference of a year in either changes the result by dozens of points.
That sensitivity is one reason modern tests switched to deviation scores, which compare a person with others of the same age. Vos Savant herself has said that even professional IQ tests are primitive measures of intelligence, a view many psychologists would share at this height.
Source: our article on the highest IQ in the world, which lists the original reports.
A confirmed 167 means a supervised extended-scale result in the top 0.0004% of your age group, in the "exceptionally gifted" band. It is far above the bars of Mensa, Intertel and the Triple Nine Society, though still short of the Mega Society's one in a million.
Levels and ranges from Gross, M. U. M. (2000), "Exceptionally and Profoundly Gifted Students: An Underserved Population", Understanding Our Gifted. They are a guide for education, not a ranking of people.
| IQ | Percentile | Rarity | Level |
|---|---|---|---|
| 163 | 99.9986th | 1 in 74,930 | Exceptionally gifted |
| 165 | 99.99926th | 1 in 136,176 | Exceptionally gifted |
| 167 | 99.9996th | 1 in 251,738 | Exceptionally gifted |
| 169 | 99.99978th | 1 in 473,383 | Exceptionally gifted |
| 171 | 99.99988th | 1 in 905,541 | Exceptionally gifted |
Yes. 167 is a very high IQ: about 1 person in 251,738 scores this high or higher, which puts it at the 99.9996th percentile. It is above the 130 that most definitions use for a high IQ, and at or above the 146 used by the Triple Nine Society for its 99.9th percentile entry.
There is no official genius score. Lewis Terman's early Stanford-Binet labels put scores over 140 in a "near genius or genius" group, but modern tests dropped the word. 167 is very rare, yet achievements, not test scores, are what usually earn the word genius.
Some can show the number, but very few can measure it. Our own test reports up to 145, because a short online test doesn't have enough very hard questions, or enough high scorers in its data, to separate 167 from nearby scores. A supervised Wechsler or Stanford-Binet assessment is needed to confirm it.
About 207. The Cattell scale uses a standard deviation of 24 instead of 15, so the same result gives a bigger number. Going the other way, a Cattell score of 167 is only about 142 on the standard scale.
An IQ of 167 is at about the 99.9996th percentile. Only about 0.0004% of people score 167 or higher.
Nobody can say exactly. Her famous 228 came from a 1937 childhood Stanford-Binet scored with the old ratio formula, and other calculations from the same records give figures from 167 to 230. Ratio scores like these aren't comparable with modern IQ scores.
Related: what counts as a good IQ score, the highest IQ in the world, famous people's IQs and the full IQ scale and range chart.
An IQ test arrives at a score like 167 by combining several kinds of task. Learn more about spatial reasoning, verbal reasoning, processing speed, fluid vs crystallized intelligence and abstract reasoning, or try the free pattern recognition test.
Many famous IQ figures are guesses. See what is really known about Terence Tao's IQ, Einstein's IQ, Isaac Newton's IQ, Nikola Tesla's IQ, Stephen Hawking's IQ, Bill Gates' IQ, Elon Musk's IQ and Garry Kasparov's IQ.
Try the free Assessivo IQ test: 36 questions, about 25 minutes, and an estimated score up to 145. Confirming a result as high as 167 needs a supervised professional test.
Take the free IQ testPortraits are pencil-style sketches by Assessivo, based on: Thomas Babington Macaulay after Sir Francis Grant, public domain. Samuel Taylor Coleridge after Washington Allston (1814), public domain.