IQ 140: how rare the score is and how to interpret it

0 The short answer

IQ 140 is approximately the 99.62nd percentile on a scale with mean 100 and standard deviation 15. Under a normal distribution, about 0.383 percent score at or above that point, roughly 1 in 261. Those are our calculations, not an observed count of people tested. The result is well above average, but it does not by itself establish genius, guarantee an outcome, or describe every cognitive domain. Check the test, edition, administration, reference group, and uncertainty before interpreting it. For other scores, use the score to percentile table; this page concerns the specific meaning of 140.

1 How is the percentile of 140 calculated?

The calculation first expresses the score in standard deviation units, then finds its position on the normal curve. Subtract the scale mean from the score and divide by the scale standard deviation. For 140 on the conventional SD 15 scale, the calculation is (140 minus 100) divided by 15, which equals 2.6667. This standardized value is usually called z.

The cumulative standard normal probability at z equals approximately 0.99617. Multiplying by 100 gives a percentile of approximately 99.617. Subtracting that cumulative probability from one gives approximately 0.00383 in the upper tail. The reciprocal of that tail probability is about 261.1. Rounding these values produces 99.62nd percentile and 1 in 261.

The standard deviation 15 explanation describes the unit of the calculation. It is a scale convention, not a property of the individual. A 40 point distance has meaning only after the unit is specified. On a scale with a wider standard deviation, the same numerical distance is fewer standard deviations from the mean.

A publisher's displayed percentile can be rounded to a whole number or derived from its own norm tables. That is not automatically an error. A theoretical normal calculation and an empirical norm table answer related questions through different procedures. Use the percentile printed on a valid report when interpreting that report; use the calculation here to understand its approximate scale position.

The IQ percentile chart gives a broader comparison across scores, and the percentile calculator makes the scale assumption explicit. Neither tool can establish that the input score was validly measured. Arithmetic can translate a score under a model, but it cannot supply the missing assessment that would make the number meaningful.

2 Does one in 261 mean exactly one person in every group?

One in 261 is an expected upper tail frequency under a model, not a quota that every group must contain. A random group of 261 independent people from the modeled population would have an expected count of approximately one person at or above 140. Expected counts need not occur exactly in a particular sample. Some such groups would have none, some one, and some more than one.

Using the same model, the probability of at least one person at or above the threshold in a group of size n is one minus (one minus p) raised to n, where p is approximately 0.00383. For n equal to 261, that is about 63 percent. This is our illustrative binomial calculation. It demonstrates why an expected count of one is different from certainty that one will be present.

Real classrooms, workplaces, universities, and hobby communities are rarely independent random samples from a national adult reference population. Their members can be selected by education, interests, geography, age, or entry requirements. A group chosen for advanced mathematics may contain more high scoring people than a random population sample. The model alone does not tell us the number in that selected group.

There is another distinction: the frequency of 140 or higher is not the frequency of an exact reported score of 140. A continuous normal variable has zero probability at one infinitely narrow point. Tests report rounded scores, so an exact integer refers to a bin or a scoring table. Calculating that bin requires an explicit convention, such as boundaries at 139.5 and 140.5, and should not be confused with an upper tail.

Nor is one in 261 a probability that a particular person's score is correct. Rarity concerns the reference distribution; measurement confidence concerns uncertainty in the assessment. A rare score can be poorly measured, and a common score can be measured precisely. These are separate dimensions of evidence.

Use the IQ rarity calculator when the question is how unusual a score is under a stated model. Avoid turning that answer into a census, a guarantee about a room, or a statement about a person's value. The calculation is useful precisely because its assumptions and its limits can be stated clearly.

3 What happens if the scale uses SD 16 or SD 24?

The same printed number can represent different relative positions when the scale standard deviation changes. Keeping the mean at 100, a score of 140 is 2.5 standard deviations above the mean on an SD 16 scale, and about 1.667 standard deviations above it on an SD 24 scale. Those positions are less extreme than 2.667 standard deviations on SD 15.

These are calculations under the same normal model. They are not empirical equivalences between named batteries. Two instruments can differ in content, norms, ceilings, and administration as well as the reporting scale. Translating a z score changes the unit; it does not make the underlying instruments identical.

To preserve the position represented by 140 on SD 15, the corresponding number on SD 16 would be 100 plus 16 times 2.6667, approximately 142.7. On SD 24 it would be 164. That is why comparing a list of impressive IQ numbers without scale information can be misleading. A larger numeral need not represent a more extreme percentile.

Mensa's own testing FAQs emphasize that different tests use different scales. Its admission criterion is expressed through an approved percentile standard, not a universal raw IQ numeral. That is a useful model for reading any score comparison: identify what is being held constant before comparing the numbers.

Historical ratio IQ adds another complication. It was constructed from a relationship between mental age and chronological age, not simply by placing adult performance on a present day normal scale. One cannot take an old ratio IQ and mechanically treat its printed number as a current deviation IQ. The history and the norm system must be understood first.

If someone reports 140 without naming the instrument, the correct next step is not to choose a flattering interpretation. Ask which scale the report uses and whether the score is a composite, an index, a converted estimate, or an informal quiz output. Until those details are known, the number does not support the precise rarity claim stated at the top of this page.

4 Is 140 a genius IQ or a gifted score?

A score of 140 belongs to a high ability range, but genius is not a universal modern test category. Historical and contemporary labels use different purposes and boundaries. The word gifted can describe an eligibility rule, a broad talent concept, or a score band. It is not a promise of a particular achievement, and its meaning should be stated in the context where it is used. On most definitions it sits inside the gifted range of roughly 130 and above, the top two percent, which is the question our guide to gifted testing for adults is written for.

In Terman's 1916 classification, the upper category was above 140 and used the language of near genius or genius. That is a historical table from an early form of intelligence testing. It should not be converted into a modern diagnostic boundary at exactly 140. The chapter on genius IQ addresses the origin of that language.

ACIS's public score chart places 135 to 144 in its Highly Gifted band. Those are ACIS public labels. They are not a reproduction of another publisher's classifications. A change in label between systems does not automatically change the standardized position of the score when the underlying mean and standard deviation are the same.

The Stanford-Binet publisher bulletin on gifted assessment explicitly discusses how gifted categories differ between versions. That is a reason to retain the edition and the reporting method, not to combine historical categories into a single ladder. Where a public sample does not verify a particular label, consult the manual rather than filling the gap from a generic chart.

An adult can use a high score as evidence of advanced performance without adopting any of these labels. If the label is needed for an application, use the receiving organization's wording and rules. If it is for personal understanding, the actual score, percentile, interval, and profile usually say more. The separate guide to gifted adults discusses that practical distinction.

5 How much uncertainty surrounds a reported 140?

The confidence interval depends on the particular score and test; 140 does not carry a universal margin of error. A reliable composite can still vary because it samples a finite set of tasks under particular conditions. The report's uncertainty should be read as part of the result, not as an optional qualification that disappears when the number is high.

The ACIS technical manual reports an FSIQ standard error of measurement of 1.60 IQ points. A simple symmetric illustration uses 1.96 times 1.60, approximately 3.14 points, on either side of the observed score. At 140, that gives approximately 136.9 to 143.1. This is our calculation from the published parameter. It is not a claim that every publisher calculates intervals this way or that every ACIS form has the Full Scale standard error.

Pearson's WAIS-5 sample report, for example, explicitly says its confidence intervals use the standard error of estimation. Borrowing an SEM and applying a familiar formula is not a substitute for the publisher's scoring procedure. Read the method attached to the report you have.

These illustrative intervals show why a small numerical difference should not be treated as a categorical difference in minds. They do not prove two people have identical ability. Comparing scores formally can require information about errors, correlations, and the relevant decision. Overlap alone is not a complete test of whether a difference is statistically meaningful.

Nor does an interval include every possible source of uncertainty. Incorrect administration, poor population fit, substantial practice exposure, or an unvalidated score conversion can create problems beyond the quoted standard error. A narrow interval printed by software is only as useful as the evidence supporting the measurement process. Verify that process before using the precision of the display to justify confidence in the score.

The SEM example belongs to the documented version 1.4 analysis. The manual retains historical coefficients after Spatial Navigation was replaced by Layer Rotation. It should not be treated as a new validation of every current or revised configuration, or applied automatically to a shorter form. Check the actual score definition and report before transferring a published coefficient.

6 Can a 140 composite contain a much lower index?

A high composite can coexist with meaningful variation across domains, because a composite summarizes several contributions. It is not a declaration that every task or index equals 140. The scoring weights, correlations, and conversion tables determine how a battery combines information. Simply averaging reported index scores is not generally the scoring method.

Consider a hypothetical profile with especially strong verbal and fluid reasoning and a less exceptional processing speed index. Such a pattern is possible, but we cannot assign a precise Full Scale value from that description alone. We would need the relevant subtest scores and the battery's authorized scoring procedure. Inventing a convenient set of indices that supposedly averages to 140 can misrepresent the actual composite.

This is why the Full Scale IQ guide separates overall level from profile direction. A domain can be a relative weakness while remaining strong compared with the reference population. If a hypothetical index is 115, that is above the mean on an SD 15 scale, even if the person's other indices are substantially higher. The within-person comparison and the population comparison answer different questions.

The profile can help explain why some activities feel easier than others, but it is not an occupation selector. A verbal index samples particular kinds of performance; it does not assess every demand of being a lawyer, teacher, or writer. Timed scanning does not capture every aspect of working efficiently. Real tasks combine skills, knowledge, tools, motivation, and opportunities to revise.

Differences also have uncertainty. Ask whether the reported gap is larger than expected from measurement error, how frequently it occurs in a suitable comparison group, and whether it matters for the actual referral question. A visually striking chart can encourage overinterpretation because it makes modest differences look dramatic. Use the numerical and contextual information as well as the picture.

Do not conclude that scatter automatically invalidates the entire composite. Nor should the overall number erase a persistent difficulty. The sensible interpretation retains both levels of information. Read the cognitive domains guide to understand what was sampled, then connect the findings to specific questions about learning or performance. A report is more useful when it describes strengths and limitations than when it treats one attractive total as the whole person.

7 What if the test has a ceiling near 140?

A ceiling limits how much an instrument can distinguish among high performances. If several very able people obtain the maximum raw score, the instrument cannot separate them using items that none missed. A displayed number may reflect the highest supported conversion, a lower bound, or an extrapolation. These possibilities need different labels. Tests sold for scores beyond the usual clinical ceiling carry the opposite risk, because that space is dominated by unstandardized take home tests without representative norms, as the guide to high range IQ tests explains.

A test that reports 140 as its maximum has not necessarily established that every person who reaches that maximum has an underlying level of exactly 140. It may only indicate performance at or above what the available tasks can resolve. The manual should explain the ceiling, the age range, and the evidence supporting upper scores. A website that permits three digit outputs beyond its norms does not solve the measurement problem.

Ceilings can occur at different stages. An individual subtest may run out of difficult items even while the overall battery retains some differentiation. A raw score conversion may have limited resolution at a particular age. An index may have extended norms with additional assumptions. These are properties to investigate in the specific edition rather than universal limits that apply to every intelligence test.

The SB5 gifted assessment bulletin illustrates why conventional scores and extended procedures need to be distinguished. An extended score is not automatically meaningless, but its interpretation depends on the procedure and validation that support it. It should not be compared casually with a standard composite from another instrument.

There is no general rule that a score becomes worthless above 130 or 140. Equally, a very high number should not be trusted merely because a test contains difficult puzzles. Difficulty alone does not establish reliable ordering or a valid population percentile. The instrument needs an appropriate scoring and normative basis.

If the result is at a ceiling and your practical decision truly requires more resolution, discuss an appropriate next assessment with a qualified professional. If the decision does not require distinguishing 140 from a higher value, another test may add little practical value. The goal is adequate evidence for the question, not an endless search for an instrument willing to print a larger number.

8 Do old norms change the meaning of 140?

A score is referenced to the population and period used by its norms, so the date and edition belong in its interpretation. A person obtaining 140 on an older instrument was compared with that instrument's reference data. That does not automatically establish the same position under a later edition, especially if content, sampling, or administration changed.

The 2015 meta-analysis by Pietschnig and Voracek documents variation in secular intelligence test score changes across time, countries, and domains. It does not support a universal correction that can be applied to every individual score. The broader discussion of the Flynn effect explains why an average historical trend is different from a personal conversion formula.

Suppose two reports from different decades both show 140. The shared numeral does not tell us whether the same constructs, item difficulty, or population frame were used. A fair comparison begins with the test names, editions, norm dates, ages at administration, and score types. It may still be impossible to equate the results precisely without a linking study.

Avoid subtracting a fixed number of points for every decade since an old assessment. That shortcut assumes the relevant trend is constant, applies to the particular ability and population, and transfers cleanly to the individual. Those assumptions are often unstated. A rough historical discussion should be labeled as such, not presented as a corrected modern IQ.

Age norms create a separate issue. Adult performance can change with age while the reported score continues to describe position among relevant peers. A stable age adjusted number does not mean every raw skill is unchanged. Conversely, an older adult's raw speed should not be judged against a young adult table unless the instrument explicitly defines that comparison.

The page on how IQ scores are normed addresses the reference frame directly. For practical use, retain the original report and do not rewrite its score as a supposedly equivalent current value. If a current decision requires current evidence, ask the receiving professional or organization what assessment is appropriate. That is more reliable than updating an old number with an internet rule.

9 Could practice or an online quiz produce a misleading 140?

A score is interpretable only when the assessment and its administration support the intended comparison. Repeated exposure to similar tasks can change performance, and the significance of that change depends on the instrument and testing interval. A higher retest score is not automatically a broad gain in the underlying abilities that the test was designed to sample.

Research on retesting in cognitive ability assessment examines practice effects across studies. It provides a reason to report previous exposure and follow the test's administration guidance. It does not justify a universal deduction from every second score. The relevant conditions, measures, and interval matter.

For an online quiz, begin with the scoring model. Does the publisher explain who took the test during norming, how age is handled, what constructs are sampled, and how uncertainty is estimated? Does a raw score have an evidence based relationship to the reported IQ scale? A polished certificate can omit every one of these questions.

An unvalidated quiz can print 140 even if the number is generated from an arbitrary formula. That is a software output, not proof of a percentile. The mathematical calculation at the beginning of this page cannot repair it. Feeding an unsupported score into a precise calculator simply produces a precise description of an unsupported input.

Testing conditions matter too. Assistance, answer searching, interruptions, or repeated attempts can change the interpretation. For self assessment, the most useful result is a fair first performance under the stated rules. There is no benefit in maximizing the display if the process makes the comparison with the norm group less meaningful.

When reviewing several results, record the test names, dates, previous exposure, and score types. Do not select the highest number and discard the rest without explanation. Differences may reflect task coverage, reliability, norms, or conditions. If the results are important and difficult to reconcile, an appropriate professional assessment can help evaluate the discrepancy. The question is why the evidence differs, not which number feels most satisfying.

10 What does 140 predict about a person's future?

A valid high score supplies information about cognitive performance; it does not uniquely determine a life outcome. Group research can show associations with learning or achievement while leaving substantial variation among individuals. A score does not measure every relevant quality, resource, opportunity, or choice. It is especially weak as a basis for predicting a stranger's character or happiness.

The 2014 longitudinal study by Lubinski, Benbow and Kell followed adolescents selected for mathematical precocity and documented diverse adult accomplishments. It supports the relevance of early ability within that selected cohort. It does not establish the career of an adult who has just received 140 on a different assessment.

Brown, Wai and Chabris's 2021 analysis also gives little support to a general claim that higher ability routinely becomes harmful across the outcomes examined. That is different from saying that 140 guarantees success. A positive group pattern leaves room for individual setbacks and for outcomes that the study did not measure.

The most defensible individual inference stays close to the tasks. If the result is a valid broad composite, it indicates high performance on the battery relative to its reference group. If the profile includes a particularly strong domain, that provides more specific information about the sampled demands. Moving from there to salary, creativity, relationships, or ethical judgment requires additional evidence.

Do not use the score as a moral ranking. The distinction between 140 and IQ 100 is a difference on an assessment scale, not a measurement of dignity. Likewise, the neighboring discussions of IQ 135 and IQ 145 describe positions and uncertainty, not separate types of human worth. Keep the interpretation proportional to what was actually measured.

11 Does 140 qualify for Mensa, Intertel, or Triple Nine?

A qualifying percentile and acceptable documentation are separate requirements. Under the SD 15 normal model, 140 exceeds the 98th and 99th percentiles but is below the 99.9th. That arithmetic can orient a reader. Admission still depends on the organization's accepted tests, score types, administration conditions, dates, and documentation.

Intertel's admission page specifies qualifying supervised evidence at or above the 99th percentile. Triple Nine's test score table lists instrument specific thresholds and restrictions; for several conventional adult scales its listed threshold is 146. The current table should be consulted rather than assuming that a converted or unsupervised 140 is sufficient.

This table was checked against the linked organization pages on September 24, 2026. It does not certify any reader's eligibility. Rules can change, and not every report using the same scale is accepted. The guide to high IQ society requirements separates those administrative questions from general score interpretation.

If applying, keep the original report. Confirm whether the organization requires direct submission from a psychologist or testing service, a particular test edition, a minimum age at testing, or a specific composite. Do not crop a screenshot so that the test name, date, or limitations disappear. Those details are part of the evidence, not unnecessary decoration.

Admission is also optional. A person can decide that a community would be enjoyable without treating membership as validation of their worth. Someone who does not qualify under a particular rule may still have advanced abilities and worthwhile interests. The decision should concern the actual community and its requirements, rather than the idea that a numerical boundary settles a person's intellectual identity.

Reading three different reports that all say 140

Imagine three documents with the same large number on the front. The first is a professionally administered adult battery with an identified edition, a dated norm reference and a confidence interval. The second is a well documented self administered assessment with a different reference sample and different task coverage. The third is a short quiz that does not explain its norms or scoring. The visible agreement of 140 does not make the underlying evidence equivalent. Before comparing the documents, establish which score each reports, whether it is a broad composite or a narrower measure, and what use its validation evidence supports.

Now suppose the first two reports both use a mean of 100 and SD 15. That resolves the scale unit, but not all the other differences. They might sample different abilities, use different age comparisons, have different ceilings or have been completed under different conditions. They could still disagree for understandable reasons, and exact agreement could partly reflect chance. Averaging them to obtain a supposedly more precise 140 would require information about their errors and dependence. Two assessments containing overlapping material do not automatically supply twice as much independent evidence as one.

A reader can make a short comparison sheet without attempting to diagnose the difference. Record the test name, edition, date, form, broad score label, norm population, interval and any prior exposure. Then record the practical question: personal understanding, an educational referral, a clinical question or an admission requirement. Missing fields are informative because they identify what the number alone cannot answer. This approach is more useful than deciding which document is correct solely from the highest result or the most impressive certificate. The guide to detailed reports explains the role of the different score levels.

Why a threshold does not turn uncertainty into a yes or no fact

The number 140 may sit above one organization's threshold and below another's, but a decision rule and an estimate are different things. An organization can choose an operational rule using a named test, an observed score and specified conditions. That rule determines eligibility under that policy. It does not remove measurement uncertainty or establish a natural boundary between two kinds of person. A neighboring score such as 139 or 141 is interpreted on the same continuous scale, even when an administrative decision changes at one of those values.

Consider a hypothetical threshold of 140 with two valid reports, one at 139 and another at 141. Whether a recipient accepts either report depends on its actual rules, which may specify an instrument, edition, date or professional verification. It would be inappropriate for this page to invent an exception based on the confidence interval. It would also be inappropriate to infer that the second person has meaningfully greater ability simply because their observed number crosses the line. The assessment question concerns uncertainty and construct coverage; the administrative question concerns an explicitly defined policy.

This distinction also matters when someone has repeated a test several times. Selecting the first score above a threshold changes the interpretation of the selected result, especially when forms or items repeat. Reporting the full assessment history allows a qualified interpreter or receiving organization to apply its rules. It avoids treating a sequence of attempts as though the most favorable one were an independent first administration. The guide to improving IQ scores distinguishes score changes from evidence of broader cognitive change; neither topic supplies permission to disregard an assessment's retest conditions.

For personal use, a boundary can be informative without becoming the center of the interpretation. A valid 140 may support a conclusion of strong performance relative to its stated reference group. The useful next questions concern the domain profile, the purpose of testing and the uncertainty that remains. A broad score does not explain every task that feels difficult, and a difficult task does not automatically disprove the broad result. Both can be described accurately when the report is read as a set of measurements rather than a single verdict.

12 What should you do with a 140 result?

First establish what the report supports, then use it for a decision that the evidence can actually inform. Preserve the test name, edition, score type, scale, date, administration conditions, and interval. Read the profile. If the purpose is an application, ask the receiving institution whether that exact evidence is accepted. If the purpose is personal understanding, connect the findings to specific learning or work questions.

ACIS has a commercial interest in this topic because we sell an online cognitive assessment. The Full Scale form covers 20 subtests across six primary domains, with methods described in the technical manual. It is unsupervised online testing. It does not provide a clinical diagnosis, a signed psychologist's report, or automatic eligibility for a high IQ society. Those limitations remain true even when the reported score is high.

For comparison with another result, use what IQ scores mean to separate constructs, scales, and norms. If the difference is small, consider measurement uncertainty before assigning significance. If the difference is large, investigate the instruments and conditions before deciding that one has revealed the person's real identity and the other has failed.

The Standards for Educational and Psychological Testing offer a framework for judging interpretations and uses of scores. The APA ethics code describes responsibilities in professional assessment. Neither source endorses a commercial product merely because the product cites it. Their relevance here is the principle that conclusions should fit the available evidence and the intended use.

An IQ of 140 can be informative and encouraging when it comes from appropriate measurement. It does not require constant retesting, an exceptional career, or comparison with famous people whose scores may never have been documented. Use it to understand measured strengths and to make better informed choices. If the practical question concerns symptoms, disability, or formal accommodations, obtain the professional evaluation required for that purpose.

13 Sources Behind This Page

Percentiles, reciprocal frequencies, scale translations, and illustrative intervals are our arithmetic under the assumptions stated beside them. The sources support assessment context, historical terminology, research boundaries, and admission requirements.

• Terman (1916). The Measurement of Intelligence. Historical classification, not current diagnostic terminology.
• Pearson. WAIS-5 sample score report. Reporting and interval method.
• Stanford-Binet 5 Assessment Service Bulletin 3. Gifted assessment and differences between score systems.
• Pietschnig and Voracek (2015). One Century of Global IQ Gains. Perspectives on Psychological Science.
• Scharfen and colleagues (2018). Retest effects in cognitive ability assessment. Intelligence.
• Lubinski, Benbow and Kell (2014). Life paths and accomplishments four decades later. Psychological Science.
• Brown, Wai and Chabris (2021). Can You Ever Be Too Smart for Your Own Good? Perspectives on Psychological Science.
• Mensa International. Testing FAQs.
• Intertel. Join Us.
• Triple Nine Society. Accepted test scores.
• AERA, APA and NCME (2014). Standards for Educational and Psychological Testing.
• American Psychological Association. Ethics Code.

14 Frequently Asked Questions