IQ 100: the exact centre of the scale

Why 100 is the centre of the scale by definition

IQ is a relative measure with no natural units. There is no absolute quantity of reasoning that equals 100 — the number exists because test publishers take a large representative sample, set that sample's average to 100, and set one standard deviation to 15 points. Every other score, including 100, is a statement about distance from that sample average. How IQ tests work walks through the norming step.

That construction is why the scale has no ceiling or floor in principle, and why a score is only ever as good as the sample behind it. Change the reference group and every number changes with it, without anybody's reasoning having altered.

Why the Average band is twenty points wide

Most classification bands span ten points. Average spans twenty, from 90 to 109. That is not an oversight — it reflects that the curve is at its tallest in the middle, so a twenty-point band there still describes a group people recognise as "typical", where a twenty-point band out at 140 would sweep together scores that differ enormously in rarity.

It also means the Average label covers real internal variation. At 100 you are 10 points into the band with 9 above you — the classification alone does not distinguish those positions, which is why the percentile is a more informative figure than the label.

Half of everyone scores in this band

About 50% of the population falls between 90 and 109. That is the single most useful thing to know about a score here: it is the most ordinary outcome there is, not a shortfall. Around 50% of people score at or above 100, and roughly 50% score below it.

The crowding has a second effect worth knowing. Because so many people are packed into the middle of the curve, one IQ point is worth about 2.7 percentile places here — the highest exchange rate anywhere on the scale. A handful of points moves your rank a surprising distance, while the same handful of points out at 145 barely moves it at all.

100 is a fixed label on a moving quantity

Because 100 is defined as the average of the reference sample rather than measured, it stays at 100 no matter what happens to actual performance. And actual performance moved: measured scores rose through most of the twentieth century, at something in the order of three points per decade in many countries — the Flynn effect. Publishers responded by periodically re-norming, which resets the average back to 100.

The consequence is specific and often missed. Someone scoring 100 today is not performing at the level someone scoring 100 in the 1950s was performing at; sitting a test normed decades ago would produce a noticeably higher number for identical raw work. 100 marks the middle of whoever was sampled most recently, so it is a fixed label attached to a quantity that has been drifting underneath it. That is a reason to treat any score as a comparison to a particular group at a particular time, not as an absolute.

Why "exactly average" is heard as a criticism

In ordinary English, average means unremarkable — sometimes worse. In psychometrics it means the most common outcome in the population, the point the whole scale is anchored to. Those two senses collide on this page more than on any other, which is why a score of 100 disappoints people more than the arithmetic warrants.

It is worth being concrete about what 100 rules out: nothing. Half the population scores at or below it, which includes a very large number of people doing demanding, skilled, complex work. The score describes rank on a timed reasoning sample, and it is silent on training, judgement, persistence and the accumulated knowledge that most real tasks actually run on.

The same score, two different minds

A composite of 100 can be arrived at in many ways, and the routes are not equivalent. One person reaches it evenly across every subtest; another combines strong verbal reasoning with below-average processing speed; a third is strong on spatial reasoning and weaker on working memory. All three score 100, and all three think differently.

This is the main limitation of any single figure, and it bites hardest in the middle of the distribution where the largest number of people sit. The subtest profile behind the composite is the more actionable information; the composite is just its average.

100 is the one composite guaranteed to exist on every test, and also the one where the strand profile is least predictable from the total: an exactly average composite can sit on top of eight near-identical strands or on a wide spread that happens to average out.

The confidence interval around 100 is the narrowest on the scale

Precision is not uniform across a test. Items cluster where test-takers cluster, which is the middle, so the middle is where a score is most tightly estimated. Around 100 there are more items discriminating between adjacent scores than anywhere else, and the largest number of norming subjects to calibrate against. That makes 100 the most trustworthy figure the scale produces — the exact opposite of the far tail, where the same instrument reports numbers it cannot really resolve. If you want to see how good IQ measurement gets, look at what it does near the average.