Average IQ in the USA: compare the numbers and the samples behind them

0 The short answer

US normed IQ scales are conventionally centered near 100 in their reference groups. That does not mean a fresh, representative national survey has measured every American at a current mean of exactly 100. Other widely shared numbers come from historical compilations or self selected online test takers and answer different questions. This page traces the samples and transformations behind 97.50 and 101.04, and separates those estimates from representative education and adult skills surveys. None of these national summaries establishes an individual American's IQ or a fixed biological characteristic of the population.

1 Why US Norms Center Near 100

US normed IQ scales conventionally center their reference performance near 100, but that does not fix the current population mean. A deviation score interprets performance relative to a defined reference group. Norm tables typically use a center of 100 and a standard deviation of 15 for IQ composites, with age appropriate comparisons. The page on how IQ scores are normed explains the process, and the standard deviation of 15 guide explains the unit. The design anchors the scale. A fresh sample of Americans can average above or below that anchor because sampling uncertainty, population coverage, cohort change and administration conditions remain empirical questions. A carefully matched norm sample is an estimate of its target population, not the population itself.

The American instruments are explicit about their samples. Pearson's Q-interactive technical report on the WAIS-5, published in September 2024 and describing a study carried out as part of the national standardization, reports a sample of nonclinical examinees recruited to be representative of the United States English speaking population aged 16 to 90 with respect to education level, race and ethnicity, and sex, according to 2022 Census Bureau data. The By the Numbers sheet gives the composition of the 2,020 examinees, 61.8 percent White, 15.4 percent Hispanic, 12.2 percent African American, 5.4 percent Asian and 5.3 percent other, and lists 20 subtests, 15 retained from the WAIS-IV and 5 new. The product page adds that the Full Scale IQ is built from seven subtests in about 45 minutes, and that the instrument is sold at qualification level C, with kits starting at 215 dollars on September 20, 2026, a price that can change. The page on what the WAIS-5 is walks through its structure.

The other clinical instrument does the same with an older census. PRO-ED's product page for the Stanford-Binet Intelligence Scales, Fifth Edition, read the same day, gives a publication date of 2003, an age range of 2 to 85 and above, and reliabilities of .95 to .98 for the Full Scale, Verbal and Nonverbal IQs. Its normative sample of 4,800 individuals closely matches the 2000 United States census, with education level based on 1999 data. The complete kit is listed at 1,353 dollars. The page on how the Stanford-Binet 5 works covers its five factor indexes. Both instruments interpret performance relative to US reference frames with different dates and designs. An individual can score above or below the center, and a later population sample can differ from it. The dates matter for norm drift but do not establish an exact conversion between editions.

Two related search questions need the same distinction. Under an ideal symmetric normal model, the mean, median and mode coincide at 100. That describes the reference scale, not a newly measured median for every American in 2026. The score versus percentile guide explains the calculation. Age norms compare examinees with reference performance at the appropriate age; they do not force every current age group in the population to average exactly 100. The average IQ by age page discusses that distinction. When comparing with the UK evidence, identify the instrument and norms. Some international datasets use British reference values, but that convention does not fix either country's current population performance.

2 The 97.50 in Becker's Dataset: 57 Samples and a 2.5 Point Assumption

The American figure in the national IQ tables is not a measurement of Americans on an American scale but a conversion of 57 mostly childhood samples onto a scale anchored on Britain, and a fixed 2.5 point deduction from an earlier country estimate enters many row conversions. The dataset file, downloaded from View on IQ on September 20, 2026, lists the United States at 92.77 on its unweighted mean of samples, 95.94 on its quality and size weighted mean, 99.07 on its conversion of school assessment results, and 97.50 on the combined measure that averages the last two. Lynn and Becker's The Intelligence of Nations, published by the Ulster Institute for Social Research in 2019 at 436 pages, is the printed companion of the same estimates; the dataset is their maintained form. The file also records the two earlier values that the older literature quotes: 98 in Lynn and Vanhanen's 2002 table and 97.5 in their 2012 revision.

The samples are what they are. Our reading of the selection sheet finds 57 American rows totaling 62,296 people: 32 scored on Wechsler scales, 16 on Stanford-Binet editions, 7 on Raven's matrices and 2 on the Kaufman battery. They were tested between 1945 and 2013, with a median year of 1981; 42 of the 57 were tested before 1990 and only 7 in 2000 or later. Their mean age is 12.4 years, 43 of the 57 have a mean age under 16, and only 9 have a mean age of 18 or above. Most are not general population samples at all. Eleven rows are labeled Black and ten White. Six are twin samples from Louisville, Texas, Boulder and unstated locations, and nine come from adoption studies in Colorado, Texas and Minnesota, counting adoptees, adoptive parents and birth mothers. The rest include Navajo and Tlingit children, Mexican American pupils, monolingual and bilingual preschoolers in New York, and the mothers of two of the child samples. The lowest corrected row is 70.5 and the highest 109.94.

Those eight rows hold 47,534 of the 62,296 people, 76 percent of the American total by our arithmetic, and all eight are samples of children aged eight or under, or of twins. Read against the norms of the tests that produced them, the 57 samples average 99.34 unweighted and 98.6 weighted by size, our arithmetic on the file's uncorrected column. Three operations turn that into 92.77. The first is a Flynn adjustment for the gap between the year of testing and the year of the norms, which the file applies row by row and which averages minus 4.48 points across the American samples; the 1973 children scored on 1947 WISC norms lose 8.84 points each. The second is the country correction. The manual that ships with the file states the rule: a sample standardized in the United States receives 97.50 minus 100.00, that is minus 2.50 points, because Lynn and Vanhanen's 2012 table put the American national IQ at 97.5 against the British anchor of 100. Fifty of the 57 American rows carry that deduction; the seven Raven's samples, scored on British norms, carry none. The third is weighting, by sample quality and size and, because the American rows are split by race, by the ethnic composition of the population, which the manual says is applied where a national standard has not been re-estimated. The result is the 95.94.

The second half of the combined figure is the school assessment conversion, 99.07. The manual describes it as national results in PISA, TIMSS and PIRLS placed on an IQ scale with a mean of 100 and a standard deviation of 15, using British means and standard deviations for the standardization, and it lists the United Kingdom at exactly 100.00 on that component. The combined 97.50 is the plain average of 95.94 and 99.07. The same file lists the United States below Germany at 100.61, the United Kingdom at 99.66, Canada at 99.62 and Australia at 99.20, and above Ireland at 95.21, Spain at 94.02 and Italy at 93.97. The page on average IQ in India shows what the same machinery does to a country whose samples are scored against British norms with no country correction of their own, and the page comparing free and validated IQ tests explains why a mean of convenience samples is not a norm.

3 The 101.04 in the Online Table: 22,605 Self-Selected Takers

The third figure describes the Americans who chose to take one online test in 2025, and the sample tells a reader who sits online IQ tests in the United States rather than what the United States scores. The International IQ Test country ranking, read on September 20, 2026, states that it was last updated on January 1, 2026 and that its sample is 1,212,714 people who took the same test on the site during 2025. The mean of the whole pool is set at 100 with a standard deviation of 15. The United States is 18th at 101.04 from 22,605 participants, up 1.30 points from the 99.74 of the previous year, when 27,841 Americans took the test. Countries with fewer than 100 takers are excluded, and territories are rolled into their sovereign country for the main ranking; Puerto Rico, listed separately, shows 93.01 from 320 takers. The site notes that in 84.8 percent of countries the average moved by less than 2 points year on year, which describes observed movement among the site's participant groups rather than establishing stable national population means.

The ACIS page on average IQ by country uses that table because it supplies a documented table, although participation and selection can vary between countries and years, and it prints a warning beside the tool. Only people with a device, a connection, one of the offered languages and an interest in taking an IQ test appear in the sample. No country was sampled by probability, none of the figures is age normed against that country's population, and a larger participant count can reduce random fluctuation within the participant pool but does not remove selection bias. The American year on year shift of 1.30 points is larger than the gap between the United States and Canada or the United Kingdom in the same table, which is the practical meaning of that warning. The review of the International IQ Test examines the provider on its own terms, and the page on whether online IQ tests are accurate treats the general problem.

The table describes observed averages among participants in that release. Similar country means do not establish similar selection mechanisms: different mixtures of age, education, language and interest can produce similar averages. Participant count alone does not establish a one point stability bound, and selection bias is not removed by a large sample. Generalizing to the US population requires evidence beyond the country table.

4 The Three Figures Side by Side

Each of the three numbers answers a different question, and the table below is the part of this page to keep, because the last column says what each figure can be used for and the first three say why.

The general lesson is the one the page on average IQ states for every population: a mean of 100 is the origin of a scale, a mean of 97.5 is a distance from somebody else's origin, and a mean of 101.04 is the middle of whoever turned up. The sections that follow take the American estimates that go beyond those three numbers, starting with the states.

5 Average IQ by State: What McDaniel's Estimates Are

McDaniel's widely reproduced state table converts school achievement results; it is not a direct survey of every state's adult IQ. McDaniel's paper in Intelligence, Estimating state IQ: Measurement challenges and preliminary correlates, built the estimates from the National Assessment of Educational Progress, the federal sample survey of public school pupils that the National Center for Education Statistics runs in every state. He took the state results in reading and mathematics for grades 4 and 8 across the years available, standardized each test in each year to a national mean of 100 and a standard deviation of 15, averaged reading and mathematics across years, and defined state IQ as the mean of the two. The reliability of the resulting rank order across years was 0.99. The estimates run from 94.2 in Mississippi to 104.3 in Massachusetts, and the full list is on the ACIS page on average IQ by state.

Three of McDaniel's own caveats decide how the table may be read. The first is coverage: NAEP publishes state results for public schools only, so a state whose abler or wealthier children attend private schools or are taught at home is underestimated, and his indicator of that bias, the share of White children outside public schools, correlates minus 0.63 with the estimates. Louisiana and Mississippi, at the bottom of the table, have the largest shares in his data. The second is exclusion and coaching: states differ in the rules for exempting pupils with cognitive impairments, and pressure to raise scores can distort state tests, which he illustrates with documented coaching on the Texas assessment. The third is that the pupils are nine and thirteen year olds in public schools, not the adult population, so the estimates say nothing about the adults who move between states. The estimates correlated 0.75 with a state health index, minus 0.58 with violent crime, 0.34 with a government effectiveness rating and 0.28 with gross state product per head. McDaniel's causal models predicted the estimates from racial composition, spending per pupil and class size with a multiple correlation of 0.83, rising to 0.88 when two birth health variables were added. That is a description of what the ordering tracks rather than of what causes it.

The 50 estimates average 100.34 with a standard deviation of 2.68, our calculations on the published table; 34 lie within 3 points of 100. The endpoint difference is 10.1 points on the converted scale. For illustration only, two hypothetical normal distributions separated by 10.1 points, each with a standard deviation of 15, would overlap substantially: a draw from the lower mean distribution would exceed a draw from the higher one about 32 percent of the time. This is our model calculation, not an observed probability for residents of Mississippi and Massachusetts. It assumes the distributions and their spreads, uses school based means for hypothetical individuals, and has no validated adult sampling model behind it. The IQ percentile chart illustrates distributional positions but cannot convert achievement estimates into personal state IQ data.

6 The Flynn Effect in America: 13.8 Points From 1932 to 1978

The United States is where Flynn first measured the effect on successive standardization samples, and the American record shows both the size of the rise and why the year of the norm table matters as much as the person. Flynn's 1984 paper in Psychological Bulletin, The mean IQ of Americans: Massive gains 1932 to 1978, showed that every Stanford-Binet and Wechsler standardization sample from 1932 to 1978 established norms of a higher standard than its predecessor. From 73 studies of 7,431 subjects aged 2 to 48 who had taken two tests with different norms, he estimated a gain of 0.300 points per year, 13.8 points across the 46 years. He also noted that the gains had acted as an unrecognized confounding variable in hundreds of studies that compared scores on tests normed in different years. The mechanism is the one this page keeps returning to: each new sample performed better in raw terms than the old one, each new mean was set to 100, and a performance that earned 100 on the old table earned less on the new one.

The general size of the effect has since been measured on larger literatures. Trahan, Stuebing, Fletcher and Hiscock's meta-analysis of 285 studies with 14,031 participants put the mean gain at 2.31 points per decade, with a 95 percent interval from 1.99 to 2.64. On modern Stanford-Binet and Wechsler tests the gain was 2.93 points per decade, an estimate the authors found not consistent with the hypothesis that the effect is diminishing. Pietschnig and Voracek's century of global IQ gains, covered 271 samples, almost 4 million people and 31 countries from 1909 to 2013. It put the annual gains at 0.41 points for fluid, 0.30 for spatial, 0.28 for full scale and 0.21 for crystallized performance, found them stronger for adults than for children, and found them decreasing in more recent decades. The two meta-analyses disagree on whether the effect is slowing, and the page on the Flynn effect sets out why the answer depends on the decades and the tests included.

These studies explain why a norm date belongs with an interpretation, but they do not predict an individual WAIS-IV to WAIS-5 difference. Multiplying a historical average gain by years between publication dates produces only a hypothetical illustration. Editions can change content, weighting, norms and administration; publication dates are not necessarily collection dates. The WAIS-IV and WAIS-5 comparison examines those changes. Actual conversion requires relevant comparison evidence from the manuals. A population assessed against fixed older norms can show a changing mean even though each new edition centers its own reference scale near 100. Renorming and population change are compatible observations, rather than reasons to declare a national mean immutable.

7 The Recent Evidence of a Reversal, Stated Narrowly

The best recent American adult evidence points the other way, and it comes from a sample large enough to be precise and unrepresentative enough that its authors decline to call the result a decline in intelligence. Dworak, Revelle and Condon's 2023 paper in Intelligence, Looking for Flynn effects in a recent online U.S. adult sample, analyzed 394,378 adults aged 18 to 90, with a median age of 29, who took cognitive items from the International Cognitive Ability Resource on the SAPA Project website between 2006 and 2018. The items covered matrix reasoning, letter and number series, verbal reasoning and, from 2011, three dimensional rotation, in an overlapping set of 35 items across the full period and 60 items from 2011. Composite scores on the 35 items fell, with an average annual difference of minus 0.029 standard deviations across levels of education in the unadjusted models, and composite scores on the 60 items fell by an average of minus 0.050 standard deviations per year between 2011 and 2018; adjusted for education and gender among 18 to 60 year olds, the 60 item slope was minus 0.034 standard deviations per year.

The pattern by domain is the informative part. Matrix reasoning and letter and number series declined, verbal reasoning slopes stayed inside the authors' threshold of 0.02 standard deviations per year and were treated as no change, and three dimensional rotation rose, at an average of 0.030 standard deviations per year, with 32 of 43 age specific slopes above the threshold. The largest differences were among 18 to 22 year olds, and the declines were steeper for participants without a four year college degree. On a scale with a standard deviation of 15, the 35 item slope corresponds to about 0.44 points per year and the adjusted 60 item slope to about 0.5 points per year, our arithmetic, which over the 12 years of the study would be a shift of roughly 5 points in a population, if the sample were one.

It is not one, and the authors say so. The participants were 65.03 percent female, a third were currently attending college, the limitations section states that the sample may be unrepresentative of the United States population, and the discussion adds that the results may not equate to real gains or declines in intelligence. The items were not a normed IQ test, the scores are standardized within the pooled 13 years of data, and nobody was tested twice. What the study supports is a narrow sentence: among American adults who visited one research website between 2006 and 2018, later visitors scored lower on matrix and series items and higher on rotation items than earlier visitors of the same age and education. The representative American evidence for the same period is the Survey of Adult Skills, examined below, and it shows declines in literacy and numeracy with its own caveats about a changed method.

8 By Age and by Education: What Changes and What Does Not

Age norms center reference performance near 100 at each age, while current population age group means remain empirical questions. The WAIS-5 compares each examinee with the norm sample of their own age band, and the Stanford-Binet 5 does the same from age 2 to 85 and above, which is why a 25 year old and a 75 year old who each score 100 have each matched the middle of a different crowd. The page on whether IQ changes with age reviews the longitudinal evidence, and the general pattern is that speed and reasoning tasks tend to peak early while vocabulary and knowledge hold or rise.

The representative American figures are in the Survey of Adult Skills. The OECD's country note for the United States reports that adults aged 55 to 65 scored 17 points lower in literacy than 25 to 34 year olds, against an OECD average gap of 30 points, and that young adults aged 16 to 24 scored 261 in literacy, 248 in numeracy and 250 in adaptive problem solving, each below the OECD average for that age group. Following the same birth cohorts across cycles, Americans born between 1989 and 1996 scored 4 points lower in literacy in 2022 to 2023 than in 2012 to 2015, a change that was not significant, while the cohort aged 44 to 54 in the first cycle scored 19 points lower ten years later, a significant decline. Those are skills on the survey's scale, not IQ points, and the survey states them as aging effects mixed with cohort differences in education.

Education is associated with substantial differences in the datasets discussed here. The same OECD note reports that among adults aged 25 to 65, those with tertiary education scored 46 points higher in literacy than those with upper secondary education, against an OECD average of 33, and those with upper secondary education scored 57 points higher than those with less, against 43. The NCES highlights of the 2023 United States results report that between 2017 and 2023 average literacy scores fell at every level of educational attainment, from 238 to 213 for adults with less than a high school education, from 259 to 246 for high school graduates and from 294 to 287 for adults with more than high school. The self-selected education table on the ACIS page on average IQ by education, from 77,974 takers of the same online test in 2025, runs from 97.3 for the no diploma group to 103.06 for four year degree holders, a gradient in the same direction and with the same selection caveats as the country table. These aggregates document associations; they do not by themselves identify individual causal effects or establish how much of a difference schooling caused.

9 What Representative American Data Actually Measure

The Survey of Adult Skills provides a national probability sample with a stated design, response rate and skills scale. The 2023 Survey of Adult Skills, the OECD's PIAAC, is a household survey conducted in respondents' homes; in the United States 3,765 adults participated, an overall response rate of 28 percent, drawn to represent about 213.3 million residents aged 16 to 65 regardless of nationality, citizenship or language. American adults scored 258 in literacy, close to the OECD average, 249 in numeracy and 247 in adaptive problem solving, both below it. In literacy 28 percent scored at or below Level 1, against an OECD average of 26 percent, and 13 percent at Levels 4 or 5, against 12. In numeracy 34 percent scored at or below Level 1, against 25 percent, and 12 percent at the top levels, against 14. In adaptive problem solving 32 percent were at or below Level 1, against 29. Taken together, 23 percent of American adults scored at the two lowest levels in all three domains, against 18 percent across the OECD.

The trend and the gaps are stated with the same precision. Average results went down compared with 2012 to 2015 in literacy and numeracy, the share of low performers rose while the share of high performers stayed stable, and the gap between the highest and lowest performing adults widened. Men scored 8 points higher than women in numeracy, a significant difference, with no significant gender difference in literacy or adaptive problem solving; numeracy declined more among men, and particularly among older men. Native born adults of native born parents, 70 percent of the participating population, scored 56 points higher in literacy than foreign born adults of foreign born parents, a gap that shrinks to 33 points after accounting for other socio-demographic factors. Among employed adults, one standard deviation of numeracy proficiency was associated with wages 11 percent higher after accounting for education and other differences. The NCES page adds that the data were collected between August 2022 and June 2023, entirely on tablets, that response rates were relatively low, and that comparisons with earlier cycles should be made with caution because of the change of mode and the new basic skills items.

None of these is an IQ, and that is the point. The scales are defined across the participating countries, the levels describe what an adult at each level can do, and the sample is a probability sample with a stated response rate and stated weighting. The page on what IQ measures explains why a score describes assessed performance relative to a defined reference group, with evidence needed for its intended interpretation, and the page on reliability and validity explains what a sample has to support for each use of a score. A national average of adult skills with those properties exists for the United States. The historical and online IQ figures reviewed here are not a comparable current probability survey of US adult IQ.

10 The USA Beside the UK and Canada

The comparisons reviewed here use different samples and scales, so their orderings should be read within their methods. In Becker's file the United Kingdom stands at 99.66, Canada at 99.62 and the United States at 97.50, and the ordering is partly written into the method: the manual's country correction gives a British normed sample zero, a Canadian normed sample plus 0.40 and an American normed sample minus 2.50, each taken from Lynn and Vanhanen's 2012 table. The British rows are largely the Raven's standardization samples that define the scale, as the page on average IQ in the UK sets out, and the American rows are the converted childhood samples described above. In the online table the United States again lies below both, while Canada and the United Kingdom exchange positions: Canada 102.09 from 7,685 takers, the United Kingdom 101.57 from 7,310 and the United States 101.04 from 22,605, differences of about a point between self-selected groups whose means moved by 0.44, 1.89 and 1.30 points in a single year. The page on average IQ in Canada traces the Canadian figures, and the pages on average IQ in Australia and average IQ in Germany do the same for the two other countries that sit near the United States in both tables.

One directly relevant comparison applies two norm sets to the same clinical participants. Harrison, Armstrong, Harrison, Iverson and Lange's 2014 paper in Archives of Clinical Neuropsychology, Comparing Canadian and American normative scores on the WAIS-IV, rescored 432 Canadian postsecondary students, referred for evaluation of learning disabilities, attention problems or other mental health concerns, against the Canadian and the American norm tables of the same test. The Canadian norms yielded IQ, index and subtest scores that were systematically lower than the American norms, and the agreement between the two systems, defined as index scores within five points or in the same classification range, was between 49 and 76 percent. The direction says that the Canadian standardization sample outperformed the American one on the same items; the size of the gap in that clinical sample is not a population figure, and the authors' conclusion was about which norms a clinician should choose. The British instrument sits on the other side of the same problem: the page on average IQ in the UK documents, from Pearson's British product page, that the WAIS-IV UK kept the American norm tables after a validation study, so a British adult's clinical IQ is a position within American norms.

The representative comparison is on the skills scale. The OECD's country note for England reports adults aged 16 to 65 at 272 in literacy, 268 in numeracy and 259 in adaptive problem solving, above the OECD average on each, with 18 percent at or below Level 1 in literacy and 14 percent at Levels 4 or 5. The American figures of 258, 249 and 247, with 28 percent low and 13 percent high in literacy, are lower on every mean and worse at the bottom of the distribution. That is a difference between two probability samples on one scale, which none of the IQ tables can offer, and it is a difference in measured skills rather than in anything an IQ test was normed to report.

11 The Evidence by Design

Read by design rather than by headline, the American evidence sorts into norm samples that define the scale, conversions of other people's samples onto other people's scales, a self-selected online pool, school achievement estimates, two records of change and one representative survey of skills, and the third column is the one to keep.

The pattern is the mirror image of the British one and the same as it. Reference groups anchor the scales used by these sources. That helps explain numerical differences between clinical norms and international compilations, but it does not settle the current population mean in either country. In both countries the only population statements with a stated sample and a stated error are on the skills scale of a survey, and in both countries that statement is not an IQ.

What the Evidence Supports, Stated Narrowly

The evidence supports five different kinds of statement about US cognitive scores, each with a specific scope. First, US clinical instruments center their reference scales near 100; this does not independently establish the present population mean or median. Second, the 97.50 in Becker's dataset is a conversion of 57 mostly childhood samples with a median test year of 1981 onto a scale anchored on British norms, in which 50 rows carry a 2.5 point deduction taken from Lynn and Vanhanen's 2012 table, and the combined figure reproduces that table's value to within 0.004 of a point. Third, the 101.04 in the online table is the mean of 22,605 self-selected takers in a pool whose mean is fixed at 100, and it moved by 1.30 points in one year. Fourth, American raw performance on Stanford-Binet and Wechsler standardizations rose by 13.8 points between 1932 and 1978, later norm samples have continued to reset the center, and the recent SAPA evidence, from an unrepresentative online sample, shows matrix and series scores falling and rotation scores rising between 2006 and 2018. Fifth, the representative 2023 Survey of Adult Skills places them near the OECD average in literacy and below it in numeracy and adaptive problem solving, with a larger share at the lowest levels than the OECD average and a decline since 2012 to 2015 that its publisher reads with caution because the method changed.

The reviewed evidence does not establish a single current US population IQ by treating 100, 97.5 and 101.04 as interchangeable. One is a reference center, another a historical aggregate with modeled adjustments, and the third a participant mean. It does not support a ranking of states as a ranking of residents, because the cited estimates describe school achievement and do not supply individual adult score distributions. It does not establish one across-the-board current trend in adult IQ by merging the SAPA findings, skills surveys with methodological changes and older norm comparisons. Each supports a narrower statement about its particular tasks, populations and periods. And it does not support the inference from any of these figures to a person, which the next section takes up.

The narrow statement is more useful than the round number because it tells a reader where to look. For the population, the Survey of Adult Skills and the NAEP are the evidence, and they are on their own scales. For a person, the evidence is a score they sat, on a stated scale, with a stated error. The page on what IQ scores mean explains how to read any score with its uncertainty, and the individual question has a different answer from the national one.

12 What This Does Not Say About You

A national average does not determine an individual American's IQ or position within a distribution. Under a normal model with mean 100 and standard deviation 15, about 68 percent of scores lie between 85 and 115 and about 95 percent between 70 and 130. Those are model proportions, not a new census of American scores. McDaniel's estimates concern school achievement, Becker's figure combines selected historical records, and the online table concerns volunteers. None gives a personal result to someone merely because they live in the United States. The IQ score chart explains scale labels, and the percentile calculator translates a stated score under stated assumptions. Input validity and reference frame suitability still need to be established.

The direction of inference matters as much as the arithmetic. The Survey of Adult Skills describes 213.3 million adults through 3,765 of them, and its levels are population shares, not a score anyone holds. The SAPA trend describes visitors to a website by year of visit. The Flynn record describes standardization samples a generation apart. Each of them is a fact about a group defined by a sampling frame, and none of them becomes a fact about a reader by virtue of the reader living in the United States. The individual level correlations between test scores and outcomes are a separate literature, and a single composite summarizes a profile without describing it.

We sell ACIS, a paid online assessment, so this paragraph carries a commercial interest. An American adult choosing any assessment should inspect its intended use, norms, administration conditions and treatment of uncertainty. A qualified professional can advise on clinical or formal documentation needs. ACIS is self-administered online and is not a clinical diagnostic instrument. Its Full Scale form contains 20 subtests across six primary domains, and the technical manual documents its reference framework and score evidence. That documentation should govern the form actually taken; a coefficient from another form or version should not be silently transferred to it. ACIS is administered in English, and this page does not promise a signed clinical report or acceptance for a society, school or employer. Review current forms and prices before purchasing. A current individual assessment can provide evidence about the person tested. It does not validate a national ranking or establish an exact current American population mean.

13 Sources Behind This Page

Every figure above is traceable to one of the following, and each is linked at the point where it is used. Dataset values for the United States and the comparison countries, the sample counts, test types, years, ages and correction values were read from the NIQ dataset file and its manual downloaded from View on IQ on September 20, 2026, and the means, weights, counts and rank described as ours were computed on that file. The state mean, standard deviation and overlap probability are our arithmetic on McDaniel's published table, and the conversions of standard deviation slopes and Flynn rates into IQ points are our arithmetic on the published figures. Product details and prices were read on the publishers' pages on September 20, 2026 and will change.

• Pearson. Wechsler Adult Intelligence Scale, Fifth Edition (WAIS-5), product page with publication date, age range, completion time, qualification level and kit prices. pearsonassessments.com, read September 20, 2026.
• Pearson. WAIS-5 By the Numbers, with the size and racial and ethnic composition of the normative sample. pearsonassessments.com, 2025, read September 20, 2026.
• Raiford S E and Courville T. Equivalence of Q-interactive and Paper Administration of Cognitive Tasks: WAIS-5. Q-interactive Technical Report 16. pearsonassessments.com, September 2024.
• PRO-ED. Stanford-Binet Intelligence Scales, Fifth Edition (SB5), complete test kit product page with normative sample and reliability. proedinc.com, read September 20, 2026.
• Becker D. The NIQ dataset, version 1.3.5, released October 25, 2023, with its manual. viewoniq.org, read September 20, 2026.
• Lynn R and Becker D. The Intelligence of Nations. Ulster Institute for Social Research, London, 2019, 436 pages. ulsterinstitute.org.
• International IQ Test. Average IQ by country 2026, ranking of 2025 test takers, as used by the ACIS country ranking page. international-iq-test.com, read September 20, 2026.
• McDaniel M A. Estimating state IQ: Measurement challenges and preliminary correlates. Intelligence, 2006, volume 34, issue 6, pages 607 to 619.
• Flynn J R. The mean IQ of Americans: Massive gains 1932 to 1978. Psychological Bulletin, 1984, volume 95, issue 1, pages 29 to 51.
• Trahan L H, Stuebing K K, Fletcher J M and Hiscock M. The Flynn effect: A meta-analysis. Psychological Bulletin, 2014, volume 140, issue 5, pages 1332 to 1360.
• Pietschnig J and Voracek M. One century of global IQ gains: A formal meta-analysis of the Flynn effect (1909 to 2013). Perspectives on Psychological Science, 2015, volume 10, issue 3, pages 282 to 306.
• Dworak E M, Revelle W and Condon D M. Looking for Flynn effects in a recent online U.S. adult sample: Examining shifts within the SAPA Project. Intelligence, 2023, volume 98, article 101734.
• OECD. Survey of Adult Skills 2023: country note for the United States. oecd.org, December 2024.
• OECD. Survey of Adult Skills 2023: country note for England (United Kingdom). oecd.org, December 2024.
• National Center for Education Statistics. PIAAC: Highlights of the 2023 U.S. national results. nces.ed.gov, read September 20, 2026.
• Harrison A G, Armstrong I T, Harrison L E, Iverson G L and Lange R T. Comparing Canadian and American normative scores on the Wechsler Adult Intelligence Scale-Fourth Edition. Archives of Clinical Neuropsychology, 2014, volume 29, issue 8, pages 737 to 746.

The Standards for Educational and Psychological Testing (AERA, APA and NCME, 2014) and the APA Ethical Principles of Psychologists and Code of Conduct, Assessment standards provide the interpretive framework: evidence must support the use, the reference group and limitations must be stated, and assessment results require appropriate explanation. Citing these standards does not imply their authors endorse ACIS or another product.

14 Frequently Asked Questions