Across much of the twentieth century, people taking intelligence tests tended to outperform people of the same age who had taken comparable tests years earlier. The common shorthand is roughly three IQ points per decade. Psychologists call this generational rise the Flynn effect, although its size has always varied by country, period and type of ability.
A formal meta-analysis spanning 1909 to 2013 found gains across 271 independent samples, nearly four million people and 31 countries. Fluid, spatial, full-scale and crystallised scores did not rise at the same rate, and gains weakened in more recent decades.
Norway then supplied an unusually clear reversal. Cognitive scores rose among young men born in the 1960s, peaked for the 1975 birth cohort, and fell for those born later. The decisive finding from a 2018 study by Bernt Bratsberg and Ole Rogeberg was that the entire curve could be reconstructed by comparing brothers within the same families.
That result strongly implicates environmental change. It does not mean that genes make no contribution to differences between individual people. It means changing genes or a changing mix of parents cannot explain the rapid rise and fall of Norway’s cohort averages.
Norway had half a century of comparable testing
The researchers linked Norwegian administrative registers to cognitive-ability results from military conscription. Their main data covered complete male birth cohorts from 1962 through 1991, generally tested at ages 18 or 19.
The average converted IQ score was 99.5 for men born in 1962. It rose to 102.3 for the 1975 cohort, then declined to 99.4 for the 1989 cohort before edging up to 99.7 for 1991. An earlier study of Norwegian conscripts had already found that large postwar gains slowed and appeared to stop by the mid-1990s. The longer run of cohorts made the later downturn easier to see.
Conscripts completed speeded tests of arithmetic, word similarities and figures. Their aggregate result was recorded on a nine-point stanine scale, which Bratsberg and Rogeberg converted to the familiar IQ scale. The test and norm remained stable across the period, apart from the mathematics section changing to a multiple-choice format in the early 1990s.
Norway was not alone. Researchers also reported declining scores among Danish conscripts. Reversals in several countries challenged the tempting belief that every generation would automatically score higher than the last.
Brothers provide an unusually strong control
Ordinary cohort comparisons mix together two kinds of change. The people composing the population may change, and the conditions experienced by people may change. A theory about genes, immigration or fertility can appear plausible if lower-scoring family types simply become more common over time.
Brothers let researchers hold many family characteristics steady. They have the same parents and share much of their household background, but they can be born into different school systems, media environments, health conditions and cultural expectations. If later-born brothers score differently from earlier-born brothers in a consistent national pattern, changing family composition cannot be the main driver.
The study’s family fixed-effects model compared scored brothers while accounting for birth year and birth order. That last adjustment matters because first-born children tend to have a small average scoring advantage. A second Bayesian model addressed selection into testing, since not every eligible man received a valid score and non-coverage increased among younger brothers in later cohorts.
The full register contained 817,611 observations. Of those, 736,808 men, or 90.1 per cent, had a valid ability score. The selection-corrected two-brother analysis then used families in which the first two children were male and born in different years.
The national curve reappeared within families
The within-family estimates reproduced the rise, the 1975 turning point and the subsequent decline found in the national averages. After correcting for selection, the model recovered the full scale of the downturn. In fact, it indicated that the decline might be slightly steeper than the raw observed scores suggested.
This is the basis for the paper’s unusually firm title: both the Flynn effect and its reversal were environmentally caused. The change operated between brothers as well as between unrelated people, leaving no need for a generational shift in population genetics to produce the pattern.
The authors concluded that genes could have at most a minor role in these cohort trends. Their findings also weakened environmental explanations that mainly differ between families, including average parental education, family size and the social effects of lower-ability parents.
That claim is about change in a population average, not the origins of every individual difference. A trait can show substantial heritability within a population while its average shifts over time because the environment changes. Heritability does not make a score immune to schooling, health, nutrition, technology or culture.
Environmental is a category, not a culprit
The sibling design eliminated several explanations, but it did not identify one cause. Conditions capable of changing within a family across the years between two sons include teaching methods, curriculum, time spent reading, exposure to digital media, health and nutrition, test familiarity, and wider attitudes towards education.
Schooling is a credible part of the story because it can change measured cognitive performance. A meta-analysis of 42 data sets and more than 600,000 participants used three quasi-experimental designs and estimated that an additional year of education was associated with about one to five IQ points, depending on the design. That does not establish schooling as the source of Norway’s decline, but it demonstrates that test performance is responsive to experience.
The same principle matters when thinking about children’s learning environments more broadly. Scandinavia Standard has previously examined the case for creative work within Scandinavian education. Such examples cannot explain an IQ trend by themselves. They illustrate why “environment” is not background noise but the changing set of activities and institutions through which cognitive skills develop.
A huge sample can still have narrow boundaries
The Norwegian data were vast, but not universal. They covered Norwegian-born men with two Norwegian-born parents who were present in Norway on their eighteenth birthday. Women and immigrant families were excluded, and the conclusions do not automatically transfer to every population.
Cohorts born after 1991 were also excluded because conscription changed substantially. Fewer than half of later cohorts were invited to in-person testing after a web survey, so their scores were no longer representative in the same way.
Nor should “three points per decade” be treated as a law. A separate meta-analysis of the Flynn effect estimated an average gain closer to 2.3 points per decade across its studies, with variation by test, age and other features. Some abilities and countries have continued rising, some have levelled, and some have fallen.
The Norwegian test battery also sampled particular skills under timed conditions. A shift in those scores is evidence about performance on arithmetic, verbal-similarity and figure tasks. It is not evidence that every form of reasoning, knowledge or practical judgement moved in lockstep.
Scores changed faster than biology could
The value of the brothers study is not that it settled what intelligence is. It showed how quickly a population statistic can change while family genetics remain effectively controlled. The rise and reversal occurred over decades and even appeared between siblings, far too quickly to require a transformation of the human gene pool.
IQ tests are renormed so that the average returns to 100. The Flynn effect refers to what happens when people from different periods are compared on a common test or common scoring standard. Later cohorts once answered more items correctly; in Norway, that direction eventually reversed.
The most accurate reading is therefore both striking and limited. Norway’s rise and decline in young men’s test scores were environmental in origin, and the within-family pattern rules out prominent genetic explanations for the cohort shift. The study does not prove which environmental change was responsible, does not erase genetic influence on individual variation, and does not show that humanity as a whole is becoming less intelligent.
It leaves a sharper question. If brothers raised by the same parents can sit on opposite sides of a national turning point, what changed in the years between them? The answer will be found in classrooms, homes, health, technology and culture, not in a sudden rewriting of their family’s DNA.