The Flynn effect is the sustained rise in average IQ test scores through most of the 20th century. On modern tests the rise runs near three points per decade. In several countries the rise has slowed or reversed since the 1990s.
- James Flynn documented a 13.8 point rise in the United States between 1932 and 1978.
- A 2014 synthesis of 285 studies found 2.31 points per decade overall. The modern subset found 2.93 for Stanford-Binet and Wechsler tests.
- Gains were largest on fluid, abstract tasks and smaller on vocabulary and stored knowledge.
- Norwegian brothers born from 1962 to 1991 show the same rise and fall within families, which points to environment rather than genes.
- An old test norm overstates IQ by about 0.3 points for each year between norming and testing.
What the Flynn effect is
The Flynn effect describes a change in test scores across generations, not a change in one person with age. Each generation took IQ tests built for its time. Later generations scored higher on the same test forms. The mean of each new standardization sample was set to 100. The same raw performance therefore earned a lower IQ on a newer test and a higher IQ on an older test.
James Flynn, a political scientist who studied intelligence test scores, documented the pattern in 1984 with United States standardization samples. The name for the pattern came later. Herrnstein and Murray used the term Flynn effect in 1994 to refer to his findings. Flynn lived from 1934 to 2020 and wrote about the meaning of the rise for the rest of his life. The original rise and the later debate are summarized in the encyclopedia overview.
A test score of 100 is a norm, not a natural constant. Test makers set the mean to 100 and the standard deviation to 15 or 16 in each new sample. When they revise a test, they restandardize it on people born more recently. A person who takes an old form today will usually score higher than on the current form. The difference comes from obsolete norms, not from a change in the person. The norm math and its use in high-stakes decisions are laid out in the 2014 synthesis.
The pattern does not mean that each past generation had very low ability. A common illustration says that people scored a century ago would average near 70 on present norms, while present people would average near 130 on past norms. Read it as arithmetic about norms. It shows how far the mean moved, not that past people could not think. Flynn himself warned against reading large gains as proof of a broad rise in real intelligence.
How fast scores rose
Flynn found a 13.8 point gain in the United States between 1932 and 1978. That equals about 0.3 points per year, or about three points per decade. Later work kept finding the same central rate on modern tests. The 13.8 point finding and the yearly rate come from Flynn 1984 as reviewed in the synthesis.
The most complete synthesis covered 285 studies with 14,031 people tested on two different test normings since 1951. The mean gain was 2.31 points per decade across all tests. For 53 comparisons with modern Stanford-Binet and Wechsler tests since 1972, with 3,951 people, the mean was 2.93 points per decade. The authors found no sign that the effect was fading in that modern set. These rates and samples are the synthesis central estimates.
Gains were not equal across all skills. A separate synthesis of 271 samples with almost four million people from 31 countries found yearly gains near 0.41 points for fluid reasoning, 0.30 for spatial skill, 0.28 for full-scale IQ and 0.21 for crystallized knowledge. Gains were stronger for adults than for children and weaker in recent decades. The domain rates and the sample size come from that global synthesis as summarized in the overview.
British data show the shape clearly. Average scores of British children on Raven Progressive Matrices rose by 14 points between 1942 and 2008. Dutch conscripts gained 21 points in 30 years between 1952 and 1982. Gains centered on Raven matrices, abstract similarities tasks and other novel reasoning problems. Vocabulary, arithmetic and stored information moved less. Reasoning with new material and knowledge from learning moved apart, as our guide to fluid and crystallized intelligence explains in full.
| Domain | What the test asks | Typical gain |
|---|---|---|
| Fluid reasoning | Solve a new matrix or series with no prior facts | About 4.1 points per decade |
| Spatial skill | Rotate a shape in mind and match it | About 3.0 points per decade |
| Full-scale IQ | Average across several subtests | About 2.8 points per decade |
| Crystallized knowledge | Define a word or state a fact | About 2.1 points per decade |
Why scores rose
No single cause carries the full rise. Research points to several environmental changes that moved together. Genes change too slowly to explain three points per decade across whole countries.
Schooling is the most direct candidate. More years in school, longer exposure to abstract tasks and more practice with hypothetical questions track the skills that rose most. One classroom change matters in particular. Modern teaching asks students to sort by abstract class, such as mammal, while past teaching favored concrete use, such as humans hunt rabbits with dogs. That shift fits the large gain on similarities tasks and other abstract subtests. Education, test familiarity, information access, complex tasks and health are reviewed as leading hypotheses.
A more complex environment adds a second push. Pictures, maps, games, machines and screens train visual analysis. Jobs demand more abstract sorting and planning than a century ago. Each generation therefore meets more novel reasoning practice outside school. Better nutrition, fewer infections, less lead exposure and better child health helped more children reach their potential. Each factor is plausible, but their exact shares remain open. Treat cause lists as hypotheses, not as measured weights.
Genetic explanations fit poorly. Family composition, migration and fertility differences cannot produce a fast, worldwide rise of this size within a few decades. The Norwegian sibling result below makes the genetic account even weaker for both the rise and the fall. Social multipliers may amplify small starts. A child with a slight advantage seeks harder problems, and harder problems build more skill. The loop can turn a modest environmental push into a larger test gain.
IQ tests also sample a narrow slice of mind. A short-answer battery misses much of what people mean by smart. For the broader map of proposed abilities, read our guide to Gardner's multiple intelligences. The Flynn rise concerns scores on these batteries, not a direct measure of worth or character.
Why the rise stopped in some places
The rise did not continue everywhere. Several developed countries show a slowdown since the 1990s and a fall in some age groups and domains. The pattern differs by country, period and subtest, so read each result with its sample and date as of 2026.
Norway gives the cleanest test. Military conscription data cover 817,611 men, with 736,808 valid scores, born from 1962 to 1991. Average scores rose from 99.5 for the 1962 cohort to 102.3 for the 1975 cohort, then fell to 99.4 for the 1989 cohort. Brothers in the same family show the same rise, turning point and fall. The within-family estimate for 1962 to 1975 runs near 0.26 points per year. The selection-corrected decline for 1975 to 1991 runs near 0.33 points per year within families and 0.34 across families. Because siblings share family background, the match rules out family composition or genes as the main driver. These cohorts, scores and within-family estimates are the study central results.
Denmark and the United Kingdom show a similar turn with different timing. Danish conscripts gained near three points per decade before 1979, then less, then fell after 1998. British 14-year-olds scored lower in 2008 than in 1980, with a drop of more than two points on average and six points in the upper half, while children aged five to ten still gained. Flynn suggested that teenage culture had stagnated, but that reading remains a hypothesis. These national patterns are surveyed with primary cites in the overview.
The United States shows a mixed recent picture. An analysis of 2006 to 2018 found declines in logic, vocabulary, visual and mathematical problem solving and analogical reasoning, while spatial skill rose. Commentators cite screens, media, schooling changes and the pandemic as possible causes. Others note a ceiling effect or a change in what society rewards. A 2026 analysis argues that IQ numbers reflect what each period values, with culturally specific items and shifting demands, so a flat or falling mean need not mean a general loss of mind. The domain split and the measurement warning come from that recent analysis.
Do not turn generation labels into facts. Searches for Gen Z or Gen Alpha and the Flynn effect are common, but the opened sources give no reliable cohort IQ series by those labels. Report Norway, Denmark, the United Kingdom and the United States as separate findings. Do not claim a worldwide fall.
What it means for reading IQ scores
Old norms mislead. A test overstates IQ by about 0.3 points for each year between the year it was normed and the year it was given. A ten-year-old norm adds about three points. A twenty-year-old norm adds about six points. Schools, clinics and courts therefore need the current form. A child tested on an old form may miss help, while an adult near a legal cutoff may face a different outcome. The yearly overestimation and the cutoff cases are documented in the synthesis.
For the general rule behind this problem, read our guide to the reliability and validity of a test score. A score is only valid against its present norm. A reliable test can still mislead when the norm is old.
Cultural limits add a second warning. IQ items favor the words, facts and problem styles of the groups who built them. A word such as tap or pipe can sort children by background rather than by thought. High scores in one place and time need not transfer to another. Read a mean of 100 as standing relative to a specific standardization sample, not as a natural law.
Use the effect to sort study questions. Follow these steps in order.
- 1.Name the comparison. Ask if the question compares generations, test forms or norm years.
- 2.Check the norm dates. Note the gap in years between norming and testing.
- 3.Apply the rough math. Multiply the year gap by 0.3 to estimate inflation from old norms.
- 4.Check the domain. Expect larger Flynn shifts on fluid and abstract tasks than on vocabulary.
- 5.Choose the measurement answer. Pick norm obsolescence when an old form gives the higher score.
- 6.Reject gene or group claims. A Flynn rise or fall alone does not explain group differences.
Save the core terms as one list in Psychology World: flynn effect, reverse flynn effect, norm obsolescence, fluid intelligence, crystallized intelligence. Each term needs one sentence in your own words that states what it is. Drill that list with Flashcard, Question and Typing until you can state the 13.8 point rise, the 2.93 modern rate and the Norway within-family result without notes. For a method that keeps such lists, read our guide to memorising psychology terminology. The Question mode needs at least four terms in the list, so keep all five together.
Frequently asked questions
What is the Flynn effect in simple terms?
It is the rise in average IQ test scores across generations during the 20th century. Later generations scored higher on the same test forms, near three points per decade on modern tests. The rise reflects environmental change and old norms, not a fixed increase in worth.
Is IQ still rising?
Not everywhere. Norway, Denmark, the United Kingdom and recent United States data show a slowdown or fall in some cohorts and domains since the 1990s. Spatial skill still rose in the United States from 2006 to 2018 while other domains fell. Read each country and period separately.
Why did IQ scores rise?
Schooling, more complex work and leisure, better health and nutrition and less exposure to harm are the leading hypotheses. Gains centered on abstract reasoning, which fits schooling that trains hypothetical thought. No study fixes their exact shares, so state them as competing explanations.
Does a higher score on an old test mean higher intelligence?
No. The same performance earns a higher IQ on an older norm because the mean was set on an earlier, lower-scoring sample. Correct for about 0.3 points per year between norming and testing and use the current test form for any decision.
Sources
- 1.Trahan, L. et al (2014). The Flynn effect: a meta-analysis. Psychological Bulletin, 140(5), 1332-1360.
- 2.Bratsberg, B. and Rogeberg, O. (2018). Flynn effect and its reversal are both environmentally caused. Proceedings of the National Academy of Sciences, 115(26), 6674-6678.
- 3.Myers, E. (2023). The Flynn effect: explaining increasing IQ scores. Simply Psychology.
- 4.Jeffries, S. (2026). IQ scores are falling but, no, we are not growing more stupid. Psyche.
- 5.Wikipedia (2026). Flynn effect.