Here is a fact that unsettles everyone who meets it. Give an adult today a test standardised in 1950 and they would score very high. Put an adult from 1950 in front of a test standardised today and they would score poorly.

The phenomenon carries the name of the New Zealand researcher James Flynn, who documented it systematically in the 1980s. The size is strikingly regular: about three points of gain per decade across industrialised countries, through most of the twentieth century.

Why you never see it

Because it is continuously erased. The IQ scale is built so that the mean is always 100. When raw performance rises, publishers restandardise on a fresh sample and the mean drops back to 100 mechanically.

Put differently: your score compares you to your contemporaries, not to humanity at large. It is rank information, re-zeroed every generation.

One practical consequence is easy to miss: a score produced by an old instrument that was never restandardised is inflated. It is among the reasons psychologists insist on current editions.

What rose, and what did not

The detail is more instructive than the headline, because the rise is very uneven.

The largest gains sit on non-verbal, abstract items: matrices, figure series, reasoning over symbols with no content. Precisely the kind of item meant to depend least on culture.

Gains are much smaller, sometimes absent, on knowledge and vocabulary items — the ones intuition would link to the spread of schooling.

That result is counter-intuitive, and it is what makes the phenomenon interesting.

The explanations on offer

None is sufficient alone, and the current consensus combines them.

Familiarity with abstraction. Flynn’s own account. In the early twentieth century, categorising the world in abstract terms was not a common reflex. Ask what a dog and a rabbit have in common: “you hunt rabbits with dogs” is concrete and entirely sensible, but the test wants “they are mammals”. A century of schooling generalised the abstract reflex — which is exactly the one these tests reward.

Nutrition and health. Better infant nutrition, the retreat of disease and parasites, reduced lead exposure. These probably explain part of the gains, particularly at the lower end of the distribution.

Family size and cognitive environment. Fewer children per household, more adult attention each, an environment richer in symbolic stimulation — screens included.

Familiarity with the format itself. Answering multiple-choice questions against a clock is a learned skill, and one now taught from primary school.

And since? A plateau, perhaps a reversal

Work in several northern European countries from the 1990s onwards has reported a plateau and, in some cases, a reversal. Those findings are debated: they often rest on particular samples — conscripts, national cohorts — and how far they generalise is contested.

Proposed explanations for a decline range from changing school practice to shifting reading habits. None is established, and it is worth resisting the easy story about declining younger generations: the data do not say that with the clarity sometimes claimed for them.

What it means for your score

A score only means something with its date and its reference. “120” without saying against which population and in which year is incomplete information.

Cross-generational comparison does not work. Your grandfather’s score on a 1960s test and yours today are not on the same scale, even if the number is identical.

What a test measures is partly cultural. If a century of social change moves performance by thirty points, the item is capturing something that depends on the era — not purely an invariant biological capacity. That is the most important point, and the one most often left out.

Our method page explains how our own scale is built and why it has to be re-anchored as the site accumulates real runs.