The Flynn effect: are IQ scores still rising – or falling?

Imagine giving a group of young adults today an intelligence test exactly as it was used on their grandparents, scored with the grandparents’ norms. For much of the 20th century, the younger group would have come out clearly ahead. This steady rise in test scores is known as the Flynn effect, after the New Zealand researcher James R. Flynn, who documented it systematically. The interesting question today is whether it is still happening.

What Flynn found

In a 1987 paper, Flynn collected data from 14 nations in which the same or comparable tests had been given to different generations. The pattern was consistent: scores had risen substantially within a single generation, in some data sets dramatically. The gains were largest on tests of abstract reasoning – the kind of matrix puzzles in which you find the rule that links shapes in rows and columns – and smaller on tests of learned knowledge such as vocabulary.

That detail matters. If the improvements had come mainly from more schooling teaching more facts, you would expect the opposite pattern. Flynn concluded that the tests could not simply be measuring a fixed, inborn ability; something in people’s environment was changing the way they approached abstract problems.

A century of data in one analysis

In 2015, Jakob Pietschnig and Martin Voracek pulled together the available evidence in a formal meta-analysis: 271 independent samples with almost four million participants, tested between 1909 and 2013. Their results confirmed gains across the century, again largest for fluid reasoning (solving new problems) and smallest for crystallised abilities (using what you have learned).

They also found that the gains had become smaller in more recent decades. Whatever drove the Flynn effect, it seemed to be running out of steam in at least some parts of the world.

The Norwegian reversal

The most cited evidence for a turnaround comes from Norway, where the vast majority of young men took the same cognitive ability test – arithmetic, word similarities and figure matrices – during military conscription. Bernt Bratsberg and Ole Rogeberg analysed scores of men born between 1962 and 1991. Scores rose for those born until the mid-1970s and then began to fall.

A decline like this could have several explanations. One possibility is a change in who has children – for example, if families with lower average scores had more children. The researchers tested this directly by comparing brothers within the same family. The rise and the later decline appeared within families too. Brothers share the same parents, so changes in which families have children cannot explain the pattern. The authors concluded that both the rise and the fall are driven by environmental factors.

What about other countries?

Data from other places is patchier. A 2023 study by Elizabeth Dworak, William Revelle and David Condon looked at nearly 400,000 US adults who took an online cognitive test between 2006 and 2018. Scores on several kinds of tasks – among them matrix reasoning, verbal reasoning and letter and number series – tended to be lower in later years, while scores on three-dimensional rotation tasks went up.

This was a large, but self-selected online sample, not a representative one, and its authors are careful about how far the results can be generalised. Still, it suggests that changes in test scores do not have to move in the same direction for every kind of ability.

Why did scores rise in the first place?

There is no single agreed answer. Explanations discussed in the research include better nutrition and health, more and longer schooling, smaller families, and a world that increasingly asks people to think in abstract categories and hypotheticals. Greater familiarity with test-like tasks may also play a role. Most researchers assume several of these factors worked together, and that their influence differed between countries and decades.

The same uncertainty applies to recent declines. Changes in education, reading habits, media use or how seriously people take a test have all been proposed. So far, no single cause has been established.

What this does not mean

• It does not mean people are suddenly “getting less intelligent”. Test scores measure performance on particular tasks under particular conditions. A drop in scores on one test is not the same as a broad decline in thinking ability.
• It does not mean genes changed. The Norwegian brother comparison points to environmental causes, and genetic change across a few decades would be far too slow to explain such shifts.
• It does not apply everywhere in the same way. Some countries show continuing gains, others stagnation or decline, and many have little good data.

Why this matters for any test

The Flynn effect has a very practical consequence: a test’s reference group goes out of date. If norms are decades old, today’s test-takers may look better (or, after a reversal, worse) than they would compared with their own generation. That is why tests need to be re-normed regularly. The IQ Arion app’s own norms are still preliminary: until about 3,000 tests have been completed, its scores are estimates marked “beta” – not a clinical assessment – and after that the norms are recalculated regularly.

Matrix puzzles, where the Flynn effect was strongest, are also part of our free short test. You can try a few and see how you approach them.