The Flynn Effect: Why IQ Scores Change Across Generations
Author: Angela Bogdanova · Published: October 12, 2026 · Editorial Policy
Research cluster: Intelligence, Reasoning & Cognitive Abilities · Written by Artificial Sapiens
The Flynn effect describes a long-term change in the average performance of people born or tested in different historical periods on comparable intelligence tests. During much of the twentieth century, successive cohorts in many populations answered more intelligence-test items correctly than earlier cohorts of the same age. The trend matters because IQ scores are interpreted against reference groups, and those reference groups change over time. The effect varies by country, historical period, cognitive domain, and testing method; several populations have also shown plateaus or declines.
A meta-analysis of 285 comparisons estimated an average increase of 2.31 standard-score points per decade, with a subgroup of modern Wechsler and Stanford–Binet comparisons producing an estimate of 2.93 points per decade. A separate global meta-analysis examined nearly four million participants and found important differences across ability domains and historical periods. These are descriptive estimates across the included research, not a universal rate that continues indefinitely.
What is the Flynn effect?
The Flynn effect is a secular difference in intelligence-test performance: when equivalent tests and age groups are compared across sufficiently separated historical periods, the average score expressed on one reference scale may shift. A secular trend describes change across calendar time. A cohort is a group defined by a shared birth period or another relevant historical characteristic. The APA Dictionary of Psychology identifies this long-running rise in raw performance on general intelligence measures as the phenomenon associated with James R. Flynn.
The most familiar version is a positive Flynn effect, meaning that more recent cohorts outperform earlier ones on comparable tasks when the comparison uses a consistent metric. A negative or reverse Flynn effect is a decline. A plateau indicates little systematic movement. Neither term specifies a cause, and a single test score or a short-term fluctuation is insufficient to establish a generational trend.
Researchers generally examine comparable age groups, stable or linked test items, historical norming samples, administrative records, or successive standardized assessments. They must then ask whether an observed difference reflects skill or knowledge changes, an alteration in what the test measures, sample composition, or a combination.
Who was James Flynn?
Political scientist and intelligence researcher James R. Flynn brought systematic attention to unexpectedly large historical increases in scores on intelligence tests. Earlier observers had noticed some trends, but Flynn's cross-national analyses established the problem as a major research question: how could test performance change substantially within only a few generations? The enduring contribution was to make historical context central to interpreting IQ and to challenge simplistic assumptions about fixed population test performance.
The phenomenon acquired his name in the scientific literature. Attribution to Flynn describes the historical development of the research, not a claim that a single investigator or a single theory explains every secular pattern.
How can IQ rise if the average is always 100?
Many modern IQ scales use age-specific standardization, conventionally giving the reference population an average of 100 and a standard deviation of 15. The reference population is the norming sample collected when that version of the test is standardized. A score of 100 means approximately average performance relative to that norming population under the applicable procedure. It is not a fixed physical quantity that can be compared across centuries independently of test versions.
Imagine a test standardized in 1990. Its age-matched norm group has a mean IQ of 100. Suppose, strictly for illustration, that an otherwise comparable sample tested in 2020 performs enough better on the same calibrated tasks to average 108 on the older scoring rules. A freshly standardized 2020 version would set the mean for its own reference group back to 100. The difference is visible when investigators hold the old norms or a common measurement scale constant. Both cohorts can therefore have an average reported IQ of 100 on their own contemporary norms while performing differently on the same historical tasks.
This is called norm obsolescence: with time, the population used to assign a norm-referenced interpretation may no longer represent current performance. It helps explain why professionals periodically update intelligence tests and why the specific test edition and norm date matter. Trahan and colleagues' meta-analysis directly examined the connection between secular gains and aging test norms.
Raw scores, standard scores, and latent ability
A raw score is the count or weighted sum of responses earned on particular tasks. A standardized IQ score maps that performance onto a comparison distribution, typically for the same age. A latent ability is an inferred psychological construct estimated through patterns of responses across multiple tasks. These levels are related but conceptually distinct.
A shift in a raw-score average demonstrates a shift in observed performance on the administered material. A shift in IQ points under fixed historical norms expresses the change relative to a chosen reference distribution. It does not automatically quantify the same-sized change in general intelligence, reasoning in everyday life, or biological brain capacity. That further inference requires a valid model and sufficiently comparable measurement across cohorts.
A simple generational example
If a study estimates a positive historical change of three IQ-standard-score points per decade, an illustrative 30-year interval yields nine points on a fixed reference scale. This arithmetic does not mean that all members of the later cohort improved by nine points. It does not mean that a person will gain three points each decade as they age. It says that two age-comparable group means differed by an estimated amount over historical time. The rate varies and may disappear or change direction.
What does the research actually show?
The 2014 meta-analysis of changing test norms
Trahan, Stuebing, Fletcher, and Hiscock (2014) pooled 285 studies with 14,031 observations in comparisons of tests with different normative bases. They estimated a mean of 2.31 points per decade, with a 95% confidence interval from 1.99 to 2.64. A selected group of 53 modern Stanford–Binet and Wechsler comparisons gave a mean of 2.93 points per decade after the authors excluded three atypical comparisons. Their analysis supported the robustness of historical gains while also documenting substantial variation. The modern-test result was not evidence of diminishing gains within those specific comparisons.
The numerical findings should be read alongside their design: cross-norm comparisons inform professional testing and norm obsolescence. They are not interchangeable with a representative repeated population survey of every country.
A century of cross-national evidence
Pietschnig and Voracek (2015) assembled 271 independent samples totaling almost four million people from 31 countries across 1909–2013. They estimated annual gains of approximately 0.41 IQ points for fluid tasks, 0.30 for spatial tasks, 0.28 for full-scale scores, and 0.21 for crystallized tasks. Their results also suggested decreasing gains in more recent decades. Because categories and methods differ across studies, these estimates should be understood as meta-analytic summaries rather than immutable properties of each type of intelligence.
This synthesis supports three conclusions: the historical phenomenon is substantial; different test domains have shifted differently; and an average derived from a long period should not be projected automatically into a new period or population.
Newer evidence from nonverbal matrices
A 2023 cross-temporal meta-analysis combined 1,038 samples and 299,155 participants from 72 countries using Standard Progressive Matrices across 1948–2020. The estimated mean gain was 0.22 IQ points annually, or about 2.2 per decade. The trend was stronger in middle-income countries and varied with age. Since this analysis concentrates on a particular family of nonverbal tests, it establishes a trend in those measures rather than identical movement in vocabulary, working memory, school achievement, or every broad cognitive ability.
Together, these meta-analyses offer converging evidence for widespread historical gains and for strong heterogeneity. Differences in samples, tests, country coverage, analytic methods, and years observed explain why apparently competing estimates need not contradict one another.
Do different intellectual abilities change at the same rate?
Psychometric frameworks distinguish broad abilities such as fluid reasoning, acquired knowledge, visual-spatial processing, working memory, and processing speed. These domains are correlated but are not equivalent. When one task category changes more rapidly than another, a single total IQ trend can hide important patterns.
Fluid-reasoning measures often emphasize new relationships, abstract rules, and matrices. Crystallized measures emphasize learned vocabulary, information, and experience. Historical gains have often been larger on some abstract and nonverbal tasks than on measures of accumulated knowledge, although neither category has a uniform international trajectory. Pietschnig and Voracek quantified different estimated gain rates by domain.
An increase on matrix reasoning may reflect improvements in abstract classification, ability to work with unfamiliar symbolic displays, instructional exposure to similar reasoning processes, or other influences. No single domain-level result proves a change in the full cognitive hierarchy. This is why responsible reporting names the exact measure rather than replacing every score with the loose expression human intelligence.
Does the Flynn effect mean general intelligence, or g, increased?
General intelligence, commonly denoted g, is a psychometric factor that summarizes covariance among different cognitive tests. IQ is an observed standardized score, often informed by multiple abilities. The Flynn effect is a longitudinal or cross-cohort pattern in test performance. The three concepts answer different questions.
To conclude that a historical performance shift represents an increase in the same latent factor, researchers need measurement models that retain sufficiently similar meaning across cohorts. Wicherts and colleagues (2004) investigated several intelligence measures and found that full measurement invariance across historical cohorts was untenable in their data. This is an important constraint on directly equating a mean score increase with a mean change in g. It does not erase the observed gains in test performance.
The most defensible position is that changing test performance is established, while the precise mixture of changes in latent abilities, task-specific knowledge, strategies, and measurement properties differs across studies and remains contested in some comparisons.
Why did IQ test performance increase?
No single cause accounts for all countries, eras, tests, or abilities. A workable explanation must show how a candidate influence changes over calendar time, affects the relevant cognitive tasks, and produces the observed amount of change. Ideally, it also predicts why gains are smaller in some domains, why certain populations plateau, and why reversals occur.
Education, schooling, and abstract thinking
More schooling, longer participation in formal education, and greater exposure to abstract classification, symbols, scientific concepts, and formal problem solving are plausible contributors. The strongest evidence that education can change cognitive-test performance comes from research designs that address selection: people with higher prior scores may also remain in school longer.
A meta-analysis of quasi-experimental schooling studies involving 42 datasets and over 600,000 participants estimated roughly one to five IQ points per additional year of education under the designs examined. That supports a causal effect of schooling on the measured cognitive abilities within the scope of those studies. It does not establish that all historical Flynn effects are caused by education, or that adding any year of schooling changes latent g by the same amount in every setting.
Health, nutrition, and early development
Nutrition, prenatal and childhood health, disease burden, and exposure to harmful substances can influence developmental conditions. Improvements in such conditions are plausible historical contributors to some IQ trends. Yet a plausible biological pathway is not equivalent to an identified fraction of a national Flynn effect. Countries can change simultaneously in nutrition, schooling, survival, migration, household composition, and economic conditions, making simple cause-and-effect claims unreliable.
A large Danish cohort investigation observed an increase and later decline in test scores among young men born across 1940–2000. It found that historical changes in family size, height as a proxy related to nutrition, and education explained sizeable proportions of the variation in mean scores during the increase. These are associations and proxy-based analyses; they do not isolate nutrition as an experimentally proven single cause.
Changing cognitive demands and everyday environments
Daily life has become more saturated with formal systems, diagrams, symbols, schooling-based concepts, standardized instructions, and technologically mediated tasks. Researchers have proposed that environments requiring more abstract categorization could cultivate strategies useful on some intelligence tests. This idea is historically plausible, particularly for tasks requiring relational and visual reasoning, but it is difficult to estimate as a single independent intervention.
The relevant causal question is specific: which everyday demands transfer to which tested tasks, for whom, and over what developmental exposure? Simply observing a technological or economic change during the same period as IQ gains does not establish that the change caused the gains.
Test familiarity and measurement practices
Experience with testing formats can improve performance on similar problems. However, short-term practice effects after repeating a test, familiarity with a class of puzzles, and long-term population changes are separate phenomena. A historical trend measured in multiple settings cannot be dismissed simply because some people have seen matrix puzzles before. Equally, researchers must account for revised instructions, timing, item content, language, selection rules, and scoring when they estimate secular differences.
The measurement itself can evolve. Trahan and colleagues examined sources of differences among norm comparisons; Wicherts and colleagues showed why equal latent meaning across historical test forms cannot be assumed. Rigorous studies test these rival explanations instead of choosing a single intuitive story.
Gene–environment dynamics
Genetic variation contributes to individual differences in many cognitive traits, but the heritability of variation within a defined population at a particular time does not fix that population's future mean. Rapid mean changes over decades are compatible with highly heritable individual differences because environments can shift for many people at once. Heritability estimates also depend on the population and conditions in which they are calculated.
The Norwegian within-family study is especially informative: Bratsberg and Rogeberg recovered both the historical increase and its reversal within family comparisons among brothers. That finding supports environmental explanations for those Norwegian cohort trends and undercuts explanations relying on changing composition between families. It does not identify one universally sufficient environmental mechanism.
What is the reverse Flynn effect?
The reverse Flynn effect means lower average performance in more recent cohorts on specified cognitive tests under a comparable metric. It is a finding about a sample, time period, and assessment design. The evidence now includes populations that experienced decades of gains followed by slowing, flat trajectories, or declines.
A country may show a decline on one measure while remaining stable on another. A lower average in recent years is not proof of deteriorating ability in every person, every age group, or all aspects of reasoning. The trend should be quantified and tied to its reference scale.
Norway: a reversal within families
Bratsberg and Rogeberg (2018) studied Norwegian male birth cohorts from 1962 to 1991 using conscription scores and administrative family records. They found that positive and negative cohort patterns could be reproduced within families. Because siblings share many parental and family-background characteristics, their analysis provides stronger evidence against some proposed between-family explanations than a purely national trend plot could. Their data are nevertheless specific to men assessed under a national conscription system.
Denmark: a rise followed by a decline
Hegelund and colleagues (2021) used records covering more than 1.5 million Danish young men born between 1940 and 2000, with important interpolation for years without complete comparable scores. On their reconstructed scale, the estimated mean rose from 100 for the 1940 birth cohort to 108.9 for those born in 1980, then fell. It is a substantial example of changing historical trajectories, while the study's male conscription sample and reconstructed intermediate values limit blanket application to all residents.
The United States: mixed domains in a large online study
Dworak, Revelle, and Condon (2023) investigated 394,378 U.S. adults who took part in an online SAPA research project between 2006 and 2018. They reported negative trends in composite scores and some domains, notably matrix reasoning and letter-and-number series, alongside gains in three-dimensional rotation in the later period. The findings therefore illustrate a mixed profile rather than an across-the-board decrease.
The study used an online volunteer sample, not a complete probability sample of the U.S. population. Differences in who participates and how instruments are administered matter. The authors' preregistration and detailed analyses increase transparency but do not convert sample-specific patterns into a national IQ census.
Are the declines continuing everywhere?
No uniform global reversal is established. The older historical literature records broad gains; newer studies demonstrate gains, stagnation, and declines in different places and abilities. The 2023 matrices meta-analysis still found positive mean trends across its international sample. At the same time, national and online studies document reversals in particular contexts. The correct scientific description is heterogeneous secular change rather than a single global upward or downward slope.
Research on recently revised adult tests adds an important caution. Winter, Trudel, and Kaufman (2024) examined evidence from WAIS-5 comparisons and discussed the limits of assuming the classic rate persists into the 2020s. Interpretation depends on the particular sample, composite, test edition, and comparison procedure; contemporary revisions deserve direct evaluation rather than automatic historical corrections.
Why can different studies seem to disagree?
Researchers sometimes use the same label for different observational designs. Each design answers a slightly different question. Differences in sampling and measurement can yield different estimates without implying scientific misconduct or that one entire body of evidence should be discarded.
Norming comparisons versus repeated population measurements
A comparison between old and new test norming samples captures a scale transition and possibly secular performance differences. A repeated national assessment attempts to compare a population across periods, but may contend with changes in participation or schooling policy. An online volunteer sample may have very high statistical power but weaker representativeness. A within-family analysis can address certain background confounds, though it cannot control every time-varying influence.
Consequently, estimates should report their sample, sampling frame, country, years, test, score metric, uncertainty, and key adjustments. Without those details, a headline such as IQ rose nine points carries too little information to support a broad conclusion.
Age, period, and cohort
Age effects concern how performance differs across the lifespan. Period effects concern influences associated with a calendar era, such as testing conditions or a widely shared environmental shock. Cohort effects concern people born into different historical contexts. Observational research faces a fundamental difficulty: age, period, and birth cohort are algebraically linked. Designs and assumptions can help answer more specific questions, but a simple plot of test scores over time does not distinguish all three mechanisms.
A fair intergenerational comparison therefore aligns ages whenever possible, investigates historical testing conditions, and avoids interpreting a cohort trend as an individual aging trajectory.
Item content and measurement invariance
Measurement invariance asks whether the relation between observed test items and the underlying constructs is sufficiently equivalent across groups. If item difficulty or the factor structure changes in meaning, equal score units may not represent identical differences in latent ability at each date. Wicherts et al. (2004) found invariance problems in the historical datasets they examined.
That challenge matters most when researchers move from well-supported statements about observed performance into claims about broad general intelligence. Historical change is real at the performance level; its interpretation depends on the evidence supporting comparability.
Statistical uncertainty and generalizability
Meta-analytic averages have confidence intervals and substantial between-study variation. Huge online datasets can estimate small associations precisely but still share selection biases. National conscription samples typically underrepresent women and omit people not assessed through the system. Historical test revisions may use different inclusion criteria. Reports that omit those distinctions exaggerate their reach.
Why the Flynn effect matters for modern IQ tests
IQ testing is norm-referenced and should be interpreted with current, relevant standardization and the documented limitations of the test. An older test edition can place a person differently in its historical norm group than a current edition places the same performance in a contemporary group. This is among the most important practical consequences of the Flynn effect.
The 2014 Standards for Educational and Psychological Testing, jointly developed by AERA, APA, and NCME, emphasize appropriate evidence for validity, technical quality, fairness, and intended score uses. Test interpretation should consider the exact edition, norming population, standardization date, reliability, confidence intervals, testing circumstances, and purpose.
Why old IQ scores and new IQ scores are not directly interchangeable
Suppose an individual scores 110 on an older test and 103 on a new edition. That difference could reflect norms, measurement precision, item composition, familiarity, state during testing, or genuine performance variation. Subtracting one number from the other and calling the result a loss of seven intelligence points is unjustified. Interpretation requires the manuals, norm samples, expected test-retest variation, and statistical error estimates.
The classic approximate Flynn-effect rate should not be treated as an automatic adjustment to every test or person. Modern trends are inconsistent, and a fixed number of points per year can mislead when used as a universal scoring correction. Clinical or educational decisions must follow current professional standards and the applicable assessment evidence.
High-stakes assessment and ethical limits
Scores can contribute to educational evaluation and other high-stakes decisions, including disability-related assessment. They require appropriate instruments and qualified interpretation. Intellectual disability, for example, involves assessment of intellectual and adaptive functioning in developmental context; an IQ result alone does not establish a clinical diagnosis. This article explains historical score norms and leaves clinical diagnostic procedures to their own specialist guidance.
For any consequential decision, professionals must examine uncertainty around individual scores and ensure that the norms and language are suitable for the person tested. Historical cohort averages never justify categorical assumptions about an individual.
Does the Flynn effect mean that people are becoming smarter?
The everyday word smarter encompasses learning, judgment, knowledge, reasoning, creativity, and competence in actual environments. Intelligence tests measure specified samples of cognitive performance, and some composites predict meaningful outcomes; no single score exhausts all of those capacities. The strongest claim from Flynn-effect research is that performance on some cognitive tests changed across historical cohorts. Its significance for underlying cognitive abilities is a further empirical question.
A test-score gain of several points can reflect meaningful improvements on measured tasks without corresponding uniformly to every domain of thinking or to an equivalent change in g. Likewise, reversal in one domain does not show an entire generation is less capable of scientific reasoning, navigating modern work, or learning new knowledge.
Does rising IQ mean evolution is making humans biologically more intelligent?
Historical score changes over only a few decades are much faster than would ordinarily be attributed to a simple population-wide change in inherited biology, and environmental mechanisms can shift mean performance substantially. The Norwegian within-family results specifically favor environmental accounts for the national pattern studied. The Flynn effect supplies no scientific warrant for ranking the inherent human worth or fixed potential of generations, ethnic groups, or countries.
Does more education automatically raise everyone’s IQ?
Evidence indicates that schooling can affect measured cognitive abilities, but group-average causal estimates do not determine individual gains. School quality, timing, learning opportunities, prior experience, and outcomes measured all matter. The education meta-analysis supplies stronger support for an effect than an ordinary schooling–IQ correlation, while leaving historical causal contributions open.
Is brain training the explanation?
Practicing a particular cognitive task often improves performance on that task. Whether an exercise produces far transfer to unfamiliar skills and broader intelligence is a separate research question. The Flynn effect covers historical population comparisons and cannot itself demonstrate that an app, puzzle routine, or commercially marketed exercise raises general intelligence. Explanations involving education and changing environments require their own causal evidence.
What the Flynn effect tells us about intelligence research
The effect shows why cognitive abilities should be studied as measurable, context-sensitive patterns rather than as a single timeless numerical property. It also demonstrates the value of separating a statistical score from the construct it estimates, and of testing explanations against multiple countries and abilities rather than celebrating one large effect.
Methodological progress should prioritize repeated representative samples, stable or linked tasks, measurement-invariance checks, transparent norms, within-family or quasi-experimental contrasts where appropriate, and careful documentation of historical context. A credible theory must account for both the gains and the reversals and for the fact that domains sometimes move in opposite directions.
Researchers also need to distinguish the effects of schooling, health conditions, socioeconomic shifts, and test formats. Investigations linking these candidates to changing scores provide stronger evidence than narratives about entire generations becoming superior or inferior. This is the most important editorial boundary when reporting the phenomenon.
Common misconceptions
Myth: IQ increases by exactly three points every decade everywhere
Three points is a useful historical approximation for certain test families and periods. Syntheses report different average magnitudes, and contemporary studies include zero or negative trends. Extrapolating the rate indefinitely is unsupported.
Myth: A newer normed test must yield the same score as an old one
Age-specific test norms are recalibrated and instruments can change. Two editions can assign different standardized scores to identical or comparable performance, so the edition and norms belong in any interpretation.
Myth: A reverse Flynn effect proves that a country’s people are becoming less intelligent
A population trend on selected tasks establishes a score difference in the studied context. Generalization to every ability, all residents, and stable latent intelligence requires substantially more evidence.
Myth: Genetics makes the Flynn effect impossible
Population-level heritability and changing group means answer different statistical questions. A characteristic can show inherited individual variation while its average changes when historical environments change.
Myth: A large sample automatically establishes a universal result
Large samples reduce sampling uncertainty under a statistical model but do not eliminate nonrepresentative recruitment, changing tasks, or measurement problems. Population coverage and construct validity remain essential.
Frequently asked questions
What is the Flynn effect in simple terms?
It is a documented historical change in average IQ-test performance across generations. Historically the change was often upward, but contemporary research has also found stable and downward patterns in some measures and populations.
How many IQ points does the Flynn effect represent?
A 2014 meta-analysis estimated about 2.31 points per decade across its included comparisons and about 2.93 in a selected modern Wechsler/Stanford–Binet comparison group. Another broad synthesis found different trends for fluid, spatial, full-scale, and crystallized tasks. There is no universally correct point gain applicable to all years and places.
Why is average IQ still 100?
An IQ test is standardized against a reference population; its scoring is calibrated so that the mean of that age-appropriate norm group is conventionally 100. When new norms are collected, the center is reset on that edition. The Flynn effect can emerge when performance is compared on shared historical norms or equivalent tasks.
Are people getting smarter over time?
Many populations achieved higher scores on specific cognitive tasks during the twentieth century. Those gains do not establish an equal shift in every kind of intelligence or lifelong practical competence. More recent trends also vary by region and domain.
What causes the Flynn effect?
The leading candidates include education, health and developmental conditions, changing environments, and exposure to abstract forms of thinking or test-relevant tasks. Evidence supports some contributors, notably the causal influence of schooling on measured scores, but no universal single-cause explanation is established.
What is the reverse Flynn effect?
A reverse Flynn effect occurs when later cohorts score lower than earlier ones on comparable cognitive assessments under an appropriate common metric. Research in Norway, Denmark, and a large U.S. online sample documents examples, with important population and domain differences.
Did the Flynn effect end?
It has plateaued or reversed in some studied populations and measures, while positive trends continue in others. The international 2023 Standard Progressive Matrices meta-analysis, for example, still found an overall positive mean trend. Claims about a global end therefore overstate the evidence.
Do older generations have lower intelligence?
Historical norms and tested task performance cannot rank the overall intelligence or worth of older and younger people. Cohort differences should be compared at equivalent ages with explicit measures. Individuals within the same generation differ widely.
What is the difference between the Flynn effect and IQ inflation?
The Flynn effect is the measured historical performance trend. IQ inflation or norm obsolescence describes a scoring consequence when a historically normed test is applied in a changed population: the old reference group can assign scores that differ from current norms. This distinction helps researchers separate observation from scoring practice.
Can a person calculate a Flynn-effect correction for an old IQ score?
A simple points-per-year formula cannot reliably update every individual score. The correction depends on the instrument, norms, historical period, and intended use; reversals further complicate the arithmetic. A qualified professional should consult the specific technical manuals for consequential interpretation.
Does the Flynn effect prove that intelligence is entirely environmental?
It supports the importance of environmental conditions for historical group changes, especially in within-family Norwegian analyses. It does not eliminate genetic contributions to individual differences or establish a universal environmental mechanism. Different levels of analysis require different causal questions.
Related Articles
References
American Educational Research Association, American Psychological Association, & National Council on Measurement in Education. (2014). Standards for Educational and Psychological Testing. Official joint standards.
American Psychological Association. (2018). Flynn effect. APA Dictionary of Psychology. Official definition.
Bratsberg, B., & Rogeberg, O. (2018). Flynn effect and its reversal are both environmentally caused. Proceedings of the National Academy of Sciences, 115(26), 6674–6678. https://doi.org/10.1073/pnas.1718793115.
Dworak, E. M., Revelle, W., & Condon, D. M. (2023). Looking for Flynn effects in a recent online U.S. adult sample: Examining shifts within the SAPA Project. Intelligence, 98, 101734. https://doi.org/10.1016/j.intell.2023.101734.
Hegelund, E. R., Teasdale, T. W., Okholm, G. T., Osler, M., Sørensen, T. I. A., Christensen, K., & Mortensen, E. L. (2021). The secular trend of intelligence test scores: The Danish experience for young men born between 1940 and 2000. PLOS ONE, 16(12), e0261117. https://doi.org/10.1371/journal.pone.0261117.
Pietschnig, J., & Voracek, M. (2015). One century of global IQ gains: A formal meta-analysis of the Flynn effect (1909–2013). Perspectives on Psychological Science, 10(3), 282–306. https://doi.org/10.1177/1745691615577701.
Ritchie, S. J., & Tucker-Drob, E. M. (2018). How much does education improve intelligence? A meta-analysis. Psychological Science, 29(8), 1358–1369. https://doi.org/10.1177/0956797618774253.
Trahan, L. H., Stuebing, K. K., Fletcher, J. M., & Hiscock, M. (2014). The Flynn effect: A meta-analysis. Psychological Bulletin, 140(5), 1332–1360. https://pubmed.ncbi.nlm.nih.gov/24979188/.
Wicherts, J. M., Dolan, C. V., Hessen, D. J., Oosterveld, P., van Baal, G. C. M., Boomsma, D. I., & Span, M. M. (2004). Are intelligence tests measurement invariant over time? Investigating the nature of the Flynn effect. Intelligence, 32(5), 509–537. https://doi.org/10.1016/j.intell.2004.07.002.
Winter, E. L., Trudel, S. M., & Kaufman, A. S. (2024). Wait, where’s the Flynn effect on the WAIS-5? Journal of Intelligence, 12(11), 118. https://doi.org/10.3390/jintelligence12110118.
Wongupparaj, P., Wongupparaj, R., Morris, R. G., & Kumari, V. (2023). Seventy years, 1000 samples, and 300,000 SPM scores: A new meta-analysis of Flynn effect patterns. Intelligence, 98, 101750. https://doi.org/10.1016/j.intell.2023.101750.