Average IQ and What It Actually Measures

Nobody ever measured the number 100 on an IQ test. It was set. The average score isn't discovered, it's defined, right at the start, before the first person ever sits down to take the test.

Here is how it works, concretely. You take a large, representative sample of people, have them complete the test, and simply assign their average performance a value of 100. Everything else is counted from that 100. An IQ test doesn't measure a quantity of intelligence the way a thermometer measures temperature. It measures where you stand relative to everyone else.

That has a few surprising consequences. One is that the same performance can produce two different numbers on two different tests. Another is that the international rankings of "average national IQ" circulating online rest on very shaky ground.

100 Is a Definition, Not a Result

The original IQ was a ratio. William Stern proposed the idea in 1912: mental age divided by actual age, times one hundred. A ten-year-old with the reasoning ability of a twelve-year-old had an IQ of 120. For children it worked reasonably well. For adults it falls apart, because mental age stops climbing much after your early twenties, so a ratio IQ would mechanically drop as you aged, without you getting any duller.

So the field switched to what is called deviation IQ, championed by David Wechsler. It doesn't compare ages. It compares your performance against your peers. You take the distribution of scores in your age group, plant the center at 100, and express the spread around it with a standard deviation. On the most common Wechsler scale, that deviation is 15 points.

That is where the famous distribution comes from. Roughly two thirds of the population fall between 85 and 115. About 2% score above 130, and a similar sliver below 70. It isn't a coincidence or a discovery. It is a mathematical consequence of how the score is built.

IQ band (Wechsler)	Approximate share of population
below 70	about 2%
70-85	about 14%
85-115	about 68%
115-130	about 14%
above 130	about 2%

And here is the first twist. Even if an entire nation grew wiser overnight, its average on this scale would stay at exactly 100. The average is always recomputed from the current sample's data. The 100 is a mirror of the group, not a fixed mark on a ruler.

Where the Bell Curve Comes From

You may have heard that intelligence has a "bell-shaped distribution" in the population. Many human traits do naturally approach a normal distribution, that symmetric curve with a hump in the middle and thin tails: height, blood pressure, shoe size. IQ, though, is largely fitted to that shape.

Raw test scores, meaning the number of correct answers, don't form a tidy bell on their own. Only during norming do the raw points get rescaled so the final scale matches a normal distribution centered on 100. It is a useful convention. It lets you know that a jump from 100 to 115 means the same shift in rank as a jump from 115 to 130. But it is still a construction, not a law of nature.

You can see the effect at the ends of the curve. Because the tails are thin, gaining points at the top gets rarer and rarer. Plenty of people can move from 100 to 110. Going from 140 to 150 is another league entirely, not because it is "only ten more points," but because almost nobody lives in that stretch of the distribution.

Same Performance, Different Number

Change the standard deviation and the numbers change too, even when your performance is identical. And there is more than one scale in use.

The Cattell scale, used by British Mensa for its entrance test, works with a deviation of 24 points. Wechsler uses 15. Someone proudly tells you they have an IQ of 148, and it sounds staggering. But if that came from the Cattell scale, it corresponds to roughly 130 on the Wechsler one, which happens to be exactly the top 2% threshold Mensa requires for membership. Two different labels, the same person.

Scale	Standard deviation	Top 2% around	Where you meet it
Wechsler	15	130	clinical WAIS and WISC tests
Cattell	24	148	British Mensa entrance test
Stanford-Binet	16	132	older and newer versions

Picture a common scene. A friend sends you a screenshot from an online test: IQ 132, victory emoji attached. Before you congratulate them, ask two things. What population was the test normed on, and which scale does it use? A lot of free web tests inflate their numbers, because it serves them to have you leave happy. And they are often normed on nothing but their own visitors, which has nothing to do with a representative sample.

For any IQ number, it pays to ask what scale produced it. Without that, a value of 130 is as ambiguous as saying "I feel warm" without a thermometer. How many points did your last app IQ test hand you, and do you even know which deviation it used?

The Average That Keeps Moving

Your 100 holds only for the group the test was normed on, and only for the moment it was set. It goes stale, and noticeably so.

The Wechsler adult test has gone through a string of revisions: WAIS-R in 1981, WAIS-III in 1997, WAIS-IV in 2008, and the newest WAIS-5 in 2024. The children's WISC versions turn over on a similar cycle. The tests get renormed every ten to fifteen years or so, and each time the average is reseated at 100 on a fresh, representative sample.

Behind this sits a pattern James Flynn documented: raw test performance rose across the 20th century, on the order of three points a decade. Without updated norms, IQ would drift upward artificially, since we would be comparing new test-takers against older, weaker results. That is a whole topic of its own. Here it is enough that renorming cancels the drift out.

What is odd is that the latest revision saw the increase shrink unexpectedly. Comparing WAIS-IV and WAIS-5, the Flynn effect came out at roughly 1.2 points instead of the expected three (a 2024 study by Winter, Trudel and Kaufman in the Journal of Intelligence). In parts of the developed world the rise has slowed, and in some places it may be reversing.

The practical upshot? Your IQ from a test twenty years ago and one from today are not directly comparable, even if you answered both identically. A lower number on the newer test doesn't mean you got duller. The bar the average is counted from simply moved in the meantime.

National Average IQ: Numbers Not to Trust

Now to the rankings you see in articles like "the smartest nations on earth." Most draw on a single database assembled by Richard Lynn and Tatu Vanhanen (the book IQ and the Wealth of Nations, 2002, and later revisions). They assigned every country one average IQ figure. The trouble is how they got there.

• Many countries rested on small, unrepresentative samples, sometimes just a few dozen children from a single city.
• For states with no data at all, the authors simply estimated IQ from neighboring countries.
• Critics have shown there were no clear criteria for which studies made it into the database and which were left out.

This is not one lone skeptic's gripe. In 2020 the European Human Behaviour and Evolution Association issued a formal statement rejecting the use of this database as methodologically unsound, and called the conclusions built on it unreliable.

Where does your own country land in these tables? Usually somewhere in the upper middle, clustered with its neighbors. But given how flimsy the underlying data is, that placement is more curiosity than fact. Differences between neighboring states are often smaller than the measurement error of the individual samples the estimates were built from. When two numbers differ by two points but each carries an uncertainty of ten, the ranking between them means nothing.

There is a deeper catch than sample quality. Even if the numbers were accurate, a nation's average IQ says far more about conditions than about innate ability. Better childhood nutrition, more years of schooling, accessible healthcare, less poverty, all of it lifts test performance. That is exactly why averages in many countries climbed over the decades as living standards improved. Explaining gaps between nations by "innate intelligence" mistakes the effect for the cause.

What the Average Says About You (and What It Doesn't)

Let's go back from nations to a single person, you. What does your personal IQ number actually mean?

First, it is a position, not a property. An IQ of 115 doesn't say you have "fifteen more units of intelligence" than someone at 100. It says you are in the top sixth or so of your age group on that particular test. It is a place in line, not a volume in a bucket.

Which is why the word "above average" can be so misleading. By definition, roughly half the population sits above 100, so on its own that is no rarity at all. The difference between 101 and 110 is invisible in practice, even though both sound "above average." Genuinely rare territory only begins at the edges, where a few percent of people reach.

Second, every score carries measurement error. Take the same test three times and you will get three slightly different numbers, depending on fatigue, mood, and focus. Serious tests report the result with a confidence interval, something like "IQ 108, 95% likely between 102 and 114." A single bare number without that range fakes a precision it doesn't have.

If you are curious where you personally stand on this scale, and above all which kinds of tasks come easier to you and which don't, you can take an IQ test that shows the spread of your performance across categories alongside the overall score. The breakdown underneath is almost always more useful than the one summary number on top.

And third, neither the average nor your personal score says anything about what IQ tests leave out: perseverance, creativity, a feel for people, the drive to see things through. Two people with the same 100 can lead entirely different lives.

So the number 100 tells you less about your head than about who you were compared with, and when. Change the group, change the era, change the scale, and the 100 shifts without you lifting a finger.