Three Points a Decade

Test publishers knew about it before anybody explained it. Every few years the norms had to be adjusted, because raw scores kept rising and the average has to be held at 100 by definition. It was treated as a technical nuisance until James Flynn assembled the data across many countries and insisted that a nuisance this large required an explanation.

The rise is roughly three points per decade, sustained through the twentieth century, found everywhere the data exists. Compounded across a hundred years that is something like thirty points, which is enormous. Flynn liked to state the consequence as a reductio: scored against today's norms, the average person of 1900 would look severely deficient, and that cannot be right.

It certainly cannot be genetic. Population-level genetic change of that magnitude over three or four generations is not possible. Something in the environment did it, and the argument has been about what ever since.

The most informative clue is that the gains were not evenly distributed. They were concentrated in abstract, fluid reasoning (matrix puzzles, similarities, classification) while crystallised knowledge such as vocabulary and arithmetic barely moved. If people had simply been getting better at everything, all the subtests would have risen together. They did not, which rules out the simplest explanations immediately.

Flynn's own account is about what he called scientific spectacles. He liked to cite interviews with Russian peasants conducted in the 1930s: asked what a dog and a rabbit have in common, a respondent answers that you use a dog to hunt rabbits. That is an intelligent, practical, entirely correct answer, and it scores zero, because the test wants an abstract category. His argument is that a century of schooling and technical work trained people to treat the hypothetical question as the real one: to answer in the mode the test is asking for, which is a learned habit rather than a capacity.

The effect has stalled or reversed in several countries with good long-running data since roughly the 1990s, particularly in Scandinavia. That is as unexplained as the rise, and the same caution applies to both: an effect this size with no agreed cause attracts confident stories, and a great many have been offered without evidence.

What the whole thing establishes, regardless of which explanation eventually wins, is one thing of considerable importance. A heritable trait can shift enormously through environmental change. Height makes the same point and is easier to accept: it is highly heritable, and average height rose dramatically across the same century through nutrition, and nobody concludes from the heritability of height that feeding children is pointless. Heritability describes variation within a population under current conditions. It is silent about what happens when the conditions change, which is exactly what the twentieth century did.