A low IQ: where the scale draws its lines at 85 and 70, and what the number cannot say

0 The short answer

Low IQ is an informal description, often used for a score below 85 on the mean 100, standard deviation 15 scale; it is not a diagnosis. Under a normal model, approximately 16 percent fall below 85 and 2.3 percent below 70, our arithmetic. A result near 70 to 75 can be relevant to assessing intellectual disability, but AAIDD also requires significant limitations in adaptive behavior and developmental onset. Read a low score with its interval, test edition, reference group and conditions. An online result cannot establish the cause of a difficulty or determine clinical severity. For a clinical question, an appropriate professional assessment must examine more than the number.

1 What Low Means on a Scale With a Mean of 100 and a Standard Deviation of 15

Low is a distance below the mean of the norm sample, and a normal model with standard deviation 15 gives a theoretical proportion below each threshold; actual samples and tests need not match it exactly. A raw score is converted into a position within the reference sample's age band, expressed as a z score, and the IQ is 100 plus 15 times that z score. A z score of minus one is an 85. A z score of minus two is a 70. The page on how IQ scores are normed describes the tables that do the conversion, and the ACIS technical manual describes the adult reference frame of 3,243 English speaking records, aged 16 to 90, against which ACIS scores are placed. The percentages that follow are our arithmetic on the standard normal distribution and describe the normal model, not a guarantee about every observed sample or test norm table.

An 85 is the 16th percentile, so about one person in six scores below it. An 80 is the 9th percentile, about one in eleven. A 75 is the 5th percentile, about one in 21. A 70 is the 2nd percentile, about one in 44. A 55 is about the 0.13th percentile, one in about 740. A 40 is about the 0.003rd percentile, one in about 31,600. The rarity calculator returns the same figures for any score, and the page on the IQ percentile chart tabulates the common ones. The table below uses the ACIS public classification, whose bands below 90 are specific to the public ACIS chart, not a universal clinical classification, and gives the share of the population in each band treating each integer score as running half a point either side of itself.

The lower tail is the mirror of the upper one. A 70 is as far below the mean as a 130 is above it, and both hold about one person in 44 on their side of the curve, which is why the gifted IQ range starts at the same distance from the centre that the clinical threshold does. The sibling page on IQ levels walks the whole scale band by band. The labels are cuts on a continuous line, and a two point difference near a boundary should be considered alongside the test's measurement uncertainty, a point the section on measurement error takes up.

Is 85 Low? Is 80 Low? The Low Average and Borderline Bands

An 85 or an 80 is below the mean and does not establish intellectual disability; its significance depends on uncertainty, functioning and assessment context. An 85 sits at the 16th percentile, within the ACIS Low Average band of 80 to 89, which holds about 15.6 percent under the half-point normal-model calculation. The guide to IQ 80 covers that band, and the guide to IQ 81 places a single score inside it at about the 10th percentile. An 80 sits at the 9th percentile. A score of 80 corresponds to approximately the ninth percentile under the normal model. The actual test report and its norm tables govern interpretation of a measured result. Whether either score counts as low depends on the convention in use: everyday usage tends to call anything below the Average band low, the research literature uses 85 as the boundary of the borderline band, while clinical assessment considers scores near 70 together with other necessary evidence.

The lines are older than the deviation scale. Lewis Terman's 1916 Stanford revision, The Measurement of Intelligence, classified 80 to 90 as dullness, 70 to 80 as border-line deficiency and below 70 as definite deficiency, in the terminology of its time, which no manual uses today. Terman also cautioned that the doubtful group near the line was not marked off by definite IQ limits. His quotient was mental age divided by chronological age, a formula that fails for adults for the reasons the page on mental age explains, while modern deviation scores describe standing relative to an age group. Similar numerical boundaries across these systems do not make their scores or clinical implications identical.

Every measured score in this region carries uncertainty. For illustration, under a hypothetical symmetric error model with a standard error of 1.60 points, an 85 spans about 82 to 88 at 95 percent confidence and an 80 spans about 77 to 83, our arithmetic. The band boundary at 79.5 falls inside the interval of an 80, so a report that prints Low Average beside an 80 is printing a label that the same person could have missed on a second sitting without any change in ability. The page on converting a score to a percentile sets out why a fixed score difference changes percentile rank more near the mean than near a tail: three points from 85 to 82 move a person from the 16th to the 12th percentile, and three points from 100 to 97 move a person from the 50th to approximately the 42nd.

2 The Borderline Band, 70 to 85: Borderline Intellectual Functioning

The band from 70 to 85 has a name in the research literature, borderline intellectual functioning, a modeled population share of about 13.6 percent, and a literature describing gaps in recognition and support. One population study of the group is Hassiotis and colleagues' secondary analysis of a United Kingdom household survey of 8,450 adults, published in the Journal of Intellectual Disability Research in 2008. In total 12.3 percent of the sample fell in the DSM-IV borderline intelligence category, close to the one eighth of the population the authors expected from the curve. Compared with adults of average intelligence, the group showed significant social disadvantage and higher rates of neurotic disorders, substance misuse and personality disorders, but no significant increase in psychotic disorder. They were more likely to receive psychiatric medication and no more likely to receive talking therapies, and they used significantly more services, including emergency services. The authors' conclusion was about care rather than about the people: the disorders in question are treatable, and services should be aware of hidden morbidity in the group.

Peltopuro, Ahonen, Kaartinen, Seppälä and Närhi's systematic review in Intellectual and Developmental Disabilities in 2014 screened 1,726 citations and included 49 studies. Adults with borderline intellectual functioning held lower skilled jobs and earned less than the general population, and the review documented neurocognitive, social and mental health hardships. The risk factors reported, such as low birth weight, and the protective factors, such as education, were not specific to the group. The review's own summary is that borderline intellectual functioning is almost invisible in the field of research despite the obvious everyday problems. Wieland and Zitman's editorial in BJPsych Bulletin in 2016 traces the reason: the DSM-IV-TR defined the category by an IQ of 71 to 84, the DSM-5 removed the IQ boundaries, and the concept was left without a clear definition in the classification systems.

Gottfredson's 1997 paper Why g matters, in Intelligence, reproduced Herrnstein and Murray's tabulation of outcomes among young White adults in the National Longitudinal Survey of Youth by IQ band. In the 76 to 90 band, 16 percent lived in poverty, 35 percent had dropped out of high school, 17 percent of mothers were chronic welfare recipients and 81 percent were married by age 30, against 6, 6, 8 and 81 percent in the 91 to 110 band. Gottfredson called the band an uphill battle rather than a high risk zone, and her stated caution applies to every figure on this page: these are odds, not guarantees, and individual outcomes vary enormously within any IQ range. The guide to IQ 70 covers the lower part of the band, where the clinical threshold begins.

3 Below 70: How Intellectual Disability Is Actually Defined

Intellectual disability requires evidence about intellectual functioning, adaptive functioning and developmental onset; an IQ score alone cannot establish it. The American Psychiatric Association's public explanation describes cognitive and adaptive limitations and emphasizes clinical judgment. Scores around 70 to 75 can indicate a significant limitation in intellectual functioning, but their meaning depends on the person's broader assessment. The developmental criterion is not established merely because an adult obtains a low score today.

The AAIDD definition specifies significant limitations in intellectual functioning and adaptive behavior originating before age 22. Adaptive behavior includes conceptual, social and practical skills. These are domains to assess, not a rule that significant deficits must be demonstrated in every one of the three domains. The professional evaluates the applicable diagnostic criteria and functioning in context. A score of 68 does not establish the diagnosis by itself, and a score of 74 does not automatically exclude it.

The distinction between cognitive performance and adaptation has a practical purpose. A standardized task asks someone to perform under defined conditions. Daily functioning concerns the skills needed in the person's environment, including communication, responsibilities, safety and independence with appropriate support. The evidence overlaps without being interchangeable. An examiner may need history, interviews, standardized measures, records and information from people who know the individual across settings.

The APA's DSM-5 revision fact sheet explains why severity is based on adaptive functioning rather than an IQ band alone. Removing a rigid score from a diagnostic criterion does not remove standardized cognitive testing from assessment. It means the number should not be elevated above the functioning that the diagnosis is intended to describe. This distinction is especially relevant when an online chart uses words resembling clinical severity levels.

A reader can ask useful questions without attempting to self-diagnose. What evidence describes intellectual functioning? How was adaptive functioning assessed? When did the relevant limitations begin? Were language, culture, sensory or motor access and the validity of the test sitting considered? These questions help distinguish an evaluation from a number displayed without context. They also prevent an adult-acquired difficulty from being automatically classified as a developmental condition.

The practical response should follow the reason for concern. Persistent difficulties with learning, independence or daily activities warrant appropriate professional evaluation, whether an informal online score is low, average or unavailable. If an institution needs documentation, establish its accepted assessment process. A consumer report does not become diagnostic because its score falls near a familiar boundary, and repeated quizzes cannot supply the adaptive and developmental evidence that the original result lacks.

4 Risk Factors and Testing Conditions: What the Evidence Shows

Research links a range of developmental exposures and testing conditions with cognitive outcomes, but the designs differ and no individual cause can be inferred from a low score alone. Genetic factors play a major part in intellectual disability, and Vissers, Gilissen and Veltman's review in Nature Reviews Genetics in 2016 describes a condition of extreme clinical and genetic heterogeneity in which sequencing has improved diagnosis and identified pathways. That is a statement about diagnosed intellectual disability, not about the heritability of scores in the general population, which the page on what heritability actually means explains is a population statistic that says nothing about the cause of one person's score. The examples below include observational associations and experimental findings; they should not all be assigned the same causal strength.

Preterm birth has been studied in cohort comparisons. Kerr-Wilson, Mackay, Smith and Pell's meta-analysis in the Journal of Public Health in 2012 pooled 27 studies and 7,044 individuals, half born preterm and half at term, and found IQ 11.94 points lower among those born preterm, with a 95 percent confidence interval of 10.47 to 13.42. Every one of the 37 estimates pointed the same way, the association did not change over the 30 years of studies, and there was a linear dose response with gestational age. Lead exposure has also been studied longitudinally. Lanphear and colleagues' pooled analysis of 1,333 children from seven longitudinal cohorts, in Environmental Health Perspectives in 2005, found a 6.9 point decrement, with a confidence interval of 4.2 to 9.4, as concurrent blood lead rose from 2.4 to 30 micrograms per deciliter. The curve was steepest at the bottom: 3.9 points were lost between 2.4 and 10 micrograms per deciliter, 1.9 between 10 and 20, and 1.1 between 20 and 30, so the largest effect per unit of exposure occurred below the level that was then treated as safe.

Maternal iodine status is one nutritional exposure examined in cohort research. Bath, Steer, Golding, Emmett and Rayman's analysis of the ALSPAC cohort, in The Lancet in 2013, measured urinary iodine in stored first trimester samples from 1,040 pregnant women in the United Kingdom and classified the group as mildly to moderately deficient, with a median urinary iodine of 91.1 micrograms per litre. After adjustment for 21 confounders, children of women whose iodine to creatinine ratio was below 150 micrograms per gram were more likely to score in the lowest quartile for verbal IQ at age 8, with an odds ratio of 1.58 and a confidence interval of 1.09 to 2.30, and for reading accuracy and comprehension at age 9, with odds ratios of 1.69 and 1.54. Scores worsened across the three exposure groups in order. For undernutrition more broadly, Grantham-McGregor and colleagues' Lancet series paper of 2007 used early childhood stunting and absolute poverty as its two indicators and estimated that over 200 million children under 5 in developing countries were not fulfilling their developmental potential, with both indicators closely associated with poor cognitive and educational performance.

Poverty also acts in the moment of testing. Mani, Mullainathan, Shafir and Zhao's Poverty impedes cognitive function, in Science in 2013, tested 464 sugarcane farmers in 54 villages of Tamil Nadu before and after the 2010 harvest, when the same farmers were poor and then relatively rich. The farmers answered 4.35 items correctly on a short set of Raven's matrices before harvest and 5.45 after, took 146 seconds against 131 on a Stroop task, and made 5.93 errors against 5.16, all with 452 to 460 participants. The authors calibrated the effect as comparable with losing a full night of sleep and, by a common approximation with a mean of 100 and a standard deviation of 15, as corresponding to about 13 IQ points. That last number is the one that circulates. What the record documents is a within person shift of about one item on a short matrices set, measured on a non normed task. Wicherts and Zand Scholten's comment in the same journal argued that the laboratory experiments, run with 101 shoppers at a New Jersey mall, did not survive reanalysis without a median split on income and were affected by ceiling effects on short, easy tests; the authors replied that the interaction held with continuous income and that the harvest result was robust to learning. The published finding and the dispute over interpretation should be distinguished; the result is not a direct measurement of a 13 point change on a full IQ battery.

Related evidence requires its own exposure and study-design qualifications. The page on fluoride and IQ distinguishes exposure levels and limits of causal interpretation, while the page on breastfeeding and IQ examines confounding and study design. None of these group findings permits identifying an individual's cause from their score, and a low result is not a reason to start a supplement or change medical care without appropriate advice.

5 Why Lists of Signs of a Low IQ Are Wrong

An informal list of behaviors cannot establish an IQ or diagnose intellectual disability. Difficulty learning a task, following instructions or understanding unfamiliar language can be a reason to investigate what support is needed. It is not a direct measurement of a person's general intellectual functioning. Clinical observation can contribute valuable evidence within an assessment, but a bystander's checklist does not replace standardized measurement, history and assessment of everyday functioning.

The same observable difficulty can have different explanations. A person may not have heard the instructions, may be using a second language, may be unfamiliar with a task, or may have a specific learning difficulty. Attention, fatigue, anxiety and opportunities to learn can also matter. The pages on ADHD and IQ and dyslexia and IQ explain why difficulty in one area should not automatically be treated as evidence of low general ability. Those possibilities are questions for assessment, not alternative diagnoses to assign from a conversation.

Base rates matter, but they do not settle the issue by themselves. Knowing that most people fall above a score threshold does not prove that every behavioral indicator will be wrong most of the time. The accuracy of an indicator also depends on its sensitivity, specificity and the setting in which it is used. A checklist without those data cannot support a claimed probability. Applying a normal-curve percentage to a coworker or relative is not an adequate substitute for that missing validation.

For practical support, describe the task rather than labelling the person. Instead of assuming a low IQ, identify the instruction that was unclear, whether information was presented accessibly, and what happened when the person received a demonstration or more time. This can improve communication without waiting for an assessment and without pretending to explain the whole cause. If difficulties persist or affect independence, an appropriate professional can help determine the next step.

The same care should apply to self-observation. Struggling with a new job, forgetting details during a stressful week or taking longer than a peer on one exercise does not establish a cognitive classification. An assessment should address the actual concern. A score, when useful, is one source of information alongside the person's history and functioning, not a replacement for listening to what they are finding difficult.

6 Test Floors: Why the Lowest IQ Ever Recorded Is Not a Record

Every normed test has a floor, the lowest score its tables can print, so the lowest possible IQ is a property of the instrument and a lowest-IQ ranking is not a meaningful psychometric comparison. The Wechsler Adult Intelligence Scale, Fifth Edition, published by Pearson in 2024, reports the Full Scale IQ on a range of 40 to 160 and the five primary index scores on a range of 45 to 155, according to Pearson's own technical report on the equivalence of its digital and paper formats, by Raiford and Courville, dated September 2024. A person who answers nothing correctly on any subtest does not receive a zero. Each subtest is scored on a scale from 1 to 19 with a mean of 10 and a standard deviation of 3, a raw score of zero maps to the lowest scaled score the table holds, and the composite built from those floors is the composite floor. The page on the WAIS-5 describes the structure of the battery; the point here is that a 40 says the person performed at or below what the table can express, and nothing more precise than that.

Floors exist because norm samples run out of people. A score of 40 is four standard deviations below the mean, about the 0.003rd percentile by our arithmetic, one person in roughly 31,600, and a typical standardization sample has limited direct observations at that extreme; instruments may use additional methods or samples, which need documentation. ACIS uses the same scaled score range of 1 to 19 and the same composite scale, and its public classification labels scores from 40 to 54 as Severe Impairment and scores below 40 as Profound Impairment, as the guide to IQ 40 explains, with the explicit caution that public interpretation in that range should defer to professional evaluation. The ACIS technical manual makes a related point about percentile language: percentiles are nonlinear: a fixed score change moves more percentile points near the mean than at the extremes, while tail rarity ratios can change substantially at the ends of the scale. The manual states this for the very high range; the same arithmetic applies at the low end.

This is why the phrase lowest IQ ever recorded has no referent. Comparing supposed record lows across instruments is uninformative because the score at the floor of an instrument carries no information beyond the fact that the floor was reached, and because the severity of intellectual disability is assessed, under the DSM-5-TR, on adaptive functioning rather than on the number. The figures that circulate under the phrase, when they have any origin at all, come from the ratio scale of the early twentieth century, on which Terman's 1916 classification placed adults with the most severe disabilities below 20 or 25 by dividing a mental age by a chronological age. Those numbers are ratios of ages, not positions on a normed curve, and they cannot be placed on the modern scale at all. The lowest score a modern adult battery can print is its floor, and the meaningful statement about a person at that floor is the one the manuals make: what they can do, with what support.

7 Measurement Error at the Threshold: Atkins, Hall and Moore

The US Supreme Court's intellectual disability decisions show why a legal decision cannot always treat a measured IQ as an exact boundary. These cases concern capital punishment and should not be read as a universal rule for all legal or educational decisions. Their psychometric lesson is that the uncertainty of a score matters, particularly when a rigid threshold could determine whether further evidence is considered.

In Atkins v. Virginia, decided in 2002, the Court prohibited execution of people with intellectual disability and left states to develop appropriate procedures for enforcing the restriction. This was the constitutional starting point. It should not be described as the same holding as the later cases concerning the handling of measurement error and adaptive evidence.

In Hall v. Florida, decided in 2014, the Court rejected Florida's use of a rigid IQ cutoff of 70 to prevent further consideration of intellectual disability. The opinion discussed the standard error of measurement and used a score of 71 with a range of approximately 66 to 76 as an illustration. That example belongs to the clinical and legal discussion in the opinion; it is not an instruction to apply exactly that interval to every modern test.

In Moore v. Texas, decided in 2017, the Court addressed a score of 74 whose measurement range included lower values, along with problems in how adaptive functioning had been evaluated. It rejected the use of nonclinical factors that risked perpetuating stereotypes. The case reinforces the need to examine the clinical framework and relevant evidence rather than stop at a single observed number.

Outside those cases, an ordinary confidence interval is still a statement about a procedure and its assumptions. A simple approximation multiplies an SEM by 1.96 and places that distance around a score. Clinical instruments can use other methods or conditional estimates. The reliability and validity guide explains precision, but the actual manual and report govern an individual result. An interval also does not compensate for an invalid administration or an unsuitable comparison group.

For example, under a hypothetical symmetric error model with SEM 1.6, a score of 70 would have an approximate interval from 66.9 to 73.1. This is arithmetic illustrating uncertainty, not a clinical diagnosis or a claim that every ACIS or Wechsler score uses the same current error model. Precision figures belong to the form and edition whose evidence produced them.

When a consequential decision depends on a low score, the relevant professional or legal representative should interpret the full evidence under the applicable requirements. A consumer article can explain the distinction between an estimate and a cutoff. It cannot determine the outcome of a person's case from a number supplied without its history, report and jurisdiction.

8 Norm Drift: The Flynn Effect Moves the 70 Line

Changing norms can change a reported score, but no universal annual correction reconstructs an individual's current IQ. Flynn's 2006 paper on capital cases discussed historical comparisons between editions and the consequences of norm obsolescence near a consequential boundary. The often cited approximation of three points per decade describes a historical pattern used in that argument, not a law that applies uniformly to every ability, country, edition or time period.

The issue is easiest to understand with a hypothetical example. If the same raw performance corresponds to 74 under one reference table and 69 under another, the category changes even though the answers do not. A real revision can also change test content, administration and scoring. Those additional changes mean that two scores from different editions cannot always be interpreted as a pure measure of norm drift.

A 2003 study by Kanaya, Scullin and Ceci examined how test renorming affected classification in school records. That finding makes the assessment edition practically relevant, but it does not justify assuming that the same amount should be subtracted from every current result. The Flynn effect guide reviews evidence that the direction and magnitude of population changes vary.

For a reader with an older report, preserve the test name, edition, testing date and age. Ask whether the assessment remains appropriate for the present purpose. If a new evaluation is needed, the clinician determines the relevant procedures and interprets differences in context. Subtracting a fixed amount for each elapsed year cannot establish developmental onset, assess adaptive functioning or diagnose intellectual disability.

This is especially important near a boundary. A numerical adjustment may look precise while introducing additional uncertainty about the reference comparison and the appropriate correction. A responsible interpretation identifies what is known about the particular tests and keeps uncertainty visible. The edition belongs with the score, rather than being a detail discarded once a label has been assigned.

9 What a Low Score Changes: School, Work and Military Entry

A low score can be relevant to learning and selection decisions, but its practical consequences depend on the evidence and rules of the particular setting. Historical research on cognitive ability and work describes group associations. It does not establish a universal minimum IQ for every occupation or show that a person below an informal cutoff cannot learn a job. The job performance review explains why predictive findings should be read with their design, outcome and limitations.

An occupational sample is not a list of everyone who can succeed. If a study reports the lower quartile of applicants to a job, that percentile describes those applicants. It is not a proven floor below which successful performance is impossible. Applicant selection, educational opportunity, prior experience and the job's actual demands all matter. Treating the lower quartile as an absolute minimum would turn a distributional description into a claim the statistic cannot support.

The same principle applies to learning. Someone may benefit from clearer instructions, demonstrations, additional practice or assistive support, but the useful approach depends on the task and the person. A broad score cannot specify the best teaching method for every subject. Identify what is difficult, what the person can already do and what changes improve performance. Where a disability or support need is suspected, assessment should address functioning directly.

Military eligibility offers a concrete example of a rule that should not be translated into an IQ cutoff. As reviewed on September 24, 2026, 10 U.S. Code section 520 generally limits the share of specified active-duty entrants scoring at or above the tenth and below the thirty-first AFQT percentile to four percent, with authority to permit an increase up to twenty percent. It also addresses requirements for people without a high school diploma. Those are provisions about AFQT results and recruitment; they are not a complete statement of every service's current eligibility policy.

The comparison of IQ and ASVAB scores explains why the scales should remain distinct. A percentile from an aptitude battery cannot automatically be converted into a clinical IQ eligibility threshold by looking up the same area on a normal curve. Anyone making an enlistment decision needs the current official service requirements and their own qualifying results. An online IQ article should not declare someone eligible or ineligible from a general score.

Educational services also follow their own evaluation and legal frameworks. The clinical definition of intellectual disability is not a complete checklist for every educational accommodation or service. Eligibility can depend on jurisdiction, disability category, educational impact and the needs of the individual. A reader should establish the actual process with the relevant school or qualified professional instead of assuming that a diagnostic threshold is the only route to help.

The practical role of a low score is therefore to contribute to a question, not to close every option. It may identify a need for further assessment or help explain a pattern already documented in learning and daily functioning. Decisions should incorporate the additional evidence that the setting requires. Group averages and old occupational summaries cannot substitute for that individual process.

10 What a Low Score Does Not Change: Adaptive Skills, Personality and Worth

A low score does not measure adaptive skill, character, personality or worth, and the diagnostic manuals say so in their own terms. The AAIDD definition separates intellectual functioning from adaptive behavior because they are related but distinct aspects of functioning, and it states that limitations often coexist with strengths in a person and that functioning improves with sustained, personalized support. The DSM-5 revision assigned severity to adaptive functioning for the same reason, and the fact sheet quoted above gives the rationale: the number must not be overemphasized as the defining factor of a person's overall ability. Two people with the same Full Scale score of 68 can differ in every practical respect, one living independently and holding a job with support in a few tasks, the other needing help across all three domains, and the diagnostic system treats them differently because the score does not distinguish them.

The morbidity figures in the borderline literature are figures about services, not verdicts about people. Hassiotis and colleagues found higher rates of neurotic disorders and substance misuse in the borderline group, and their conclusion was that the disorders are treatable and that services were failing to recognize hidden morbidity; the same study found no significant increase in psychotic disorder and found that the group received medication rather than talking therapies. Peltopuro's review found that the risk and protective factors it identified were not specific to the group. On crime, the page on IQ and crime reviews the cohort evidence, its confounding by socioeconomic status and the effect of differential detection, and cautions against applying group findings to an individual.

Personality is a separate set of measurements with a separate literature, reviewed on the page on IQ and personality, and a low score does not measure or determine conscientiousness, warmth, honesty or humor. Nor does it make an adult a child. The idea that a person with an IQ of 70 has the mind of a ten year old is a leftover of the ratio quotient, which divided a mental age by a chronological age and which fails for adults because test performance stops rising in the late teens while age does not. An adult at 70 has an adult's experience, memory, relationships and judgment about their own life, and the AAIDD's list of adaptive skills, from travel and schedules to social responsibility and the ability to follow rules and obey laws, is a list of things that adults at every score do.

11 Retest and Profile: A Low Full Scale Can Hide an Average Domain

A broad composite and its domain scores provide different information, so a low total can coexist with stronger performance in a particular area. The six cognitive domains describe the performances sampled by ACIS, and the Full Scale IQ guide explains what the broad summary represents. Variation in a profile does not automatically invalidate the total. The total describes overall level across contributing tasks; the component scores can add detail about relative strengths and weaknesses. That comparison is possible only on a battery broad enough to report domains, which is the difference between a full scale battery and a quiz built on one kind of task.

For a simple illustration, imagine a verbal index of 92 and a processing speed index of 66 from the same appropriately administered battery. The 26 point gap is an observed difference. Its interpretation requires the instrument's comparison procedures, relevant error estimates and information about how common such a gap is in the reference group. The numbers alone do not establish a particular condition. They also do not permit calculating an exact Full Scale IQ without the necessary subtest scores and scoring tables.

A common shortcut calculates a standard error of the difference by taking the square root of the sum of two squared SEMs. That formula assumes uncorrelated measurement errors. It should not be applied automatically to every pair of indices or used as a substitute for the instrument's published comparison tables. Statistical significance, rarity of a difference and clinical usefulness are separate questions. A large observed gap may deserve attention without identifying its cause.

Testing conditions matter too. Timed performance may be affected by access to the task, interruptions, fatigue or other circumstances. The processing speed explanation describes those task demands, while the page on twice exceptional profiles explains why a strength and a difficulty can coexist. Neither page turns an uneven profile into a diagnosis. Interpretation should be tied to the person's history and the purpose of assessment.

Retesting is not automatically a correction of an earlier low score. Estevis, Basso and Combs studied practice effects when adults repeated the WAIS-IV after short intervals. Their average change should not be presented as the gain every individual will obtain on every retest. Practice, temporary conditions, measurement error and actual changes in performance can contribute differently in different cases.

Education can also affect measured cognitive performance. Ritchie and Tucker-Drob's meta-analysis examined quasi-experimental evidence linking additional education with higher intelligence test scores. That finding supports a population-level causal question under the studies' assumptions. It is not a personalized forecast of how many points a reader will gain from a particular course, and it does not mean that a low first score is always a lower bound on true ability.

A repeat result may be higher, lower or similar. Keep the dates, instruments, conditions and any prior exposure together when comparing. If the first score arose during an unsuitable sitting, seek an appropriate interpretation of validity rather than simply discarding the lower number. If the concern is clinical, a qualified professional can decide whether retesting, another instrument or a different kind of evidence will be most informative.

12 What ACIS Reports, and What It Cannot Diagnose

ACIS is a paid self-administered online assessment, and its scores do not diagnose intellectual disability or establish clinical severity. The technical manual describes the adult reference frame, analysis set, score construction and interpretation limits. Its stated frame includes 3,243 English speaking adult records aged 16 to 90, while the technical analysis uses 2,750 complete records. Those counts describe different roles in the documentation; they do not independently establish a sampling design.

The assessment reports cognitive performance across its included tasks. It does not administer a complete clinical evaluation of adaptive functioning and developmental history, so its output cannot establish the diagnostic framework discussed above. It should not be marketed or read as a validated screening instrument for intellectual disability merely because its numerical scale includes a lower range. A reader with persistent clinical concerns should use an assessment route designed for those concerns.

Precision claims must also retain their form and edition. The manual publishes reliability and error information, together with a notice about a September 2026 change to the visual spatial battery. Historical coefficients should not be presented as fresh validation of every replacement task or composite. For a current report, read the applicable documentation and any qualifications rather than assuming that a single SEM quoted in an article governs every result.

Unsupervised online conditions add interpretation requirements. A result obtained with interruptions, unfamiliar instructions, unsuitable access conditions or incomplete engagement may not answer the intended question. An ordinary confidence interval does not repair those problems. The reason for assessment should determine whether a self-administered result is useful, whether additional evidence is needed, or whether a professional evaluation is appropriate from the outset.

If the goal is formal documentation, establish what the receiving institution accepts before selecting a test. Requirements differ across clinical, educational and administrative settings. The page on who can administer an IQ test explains professional assessment roles. An ACIS result does not become an accepted institutional report because its number falls in a familiar band.

For personal interpretation, begin with the conditions and the relevant score report, then consider the broad result alongside the profile. A low score can prompt a useful question about performance or support, but it cannot answer every question about a person's ability, history or daily functioning. The useful outcome is a clearer understanding of what the evidence supports and what further information is needed.

What conclusions does the evidence support?

A low result describes performance relative to a reference system under particular conditions. A score near 70 requires careful attention to uncertainty, and a diagnosis requires additional adaptive and developmental evidence. Group associations can identify risks worth studying or addressing, but they do not identify the cause of one person's score. Support and practical learning should be considered in relation to the person's needs rather than deferred until a chart supplies the right label.

The next step depends on the question behind the score. Preserve the original report, check the instrument and norm context, and describe the difficulties or decisions that motivated assessment. When the concern is clinical, educational or legal, take that question to the appropriate professional or institution. A number can contribute evidence, but it should not be asked to supply an evaluation it never performed.

13 Sources Behind This Page

The sources below distinguish diagnostic criteria, observational risk research, measurement evidence and the specific legal examples. ACIS reliability and standard error figures are from the technical manual, version 1.4; normal-model proportions and hypothetical interval examples are arithmetic illustrations, not diagnoses or observed prevalence estimates. Terman's 1916 classification, Wieland and Zitman's 2016 editorial, Vissers and colleagues' 2016 review, Grantham-McGregor and colleagues' 2007 estimate, Kanaya and colleagues' 2003 study, the Estevis 2012 retest study, the Ritchie and Tucker-Drob 2018 meta-analysis and the published comment and reply on the Mani study are linked in the body at the point of use with their figures. Official definitions and the statute were reviewed on September 24, 2026; historical research dates are retained.

• American Psychiatric Association. What is intellectual disability? Diagnostic criteria, prevalence and severity as described in the DSM-5-TR. psychiatry.org, reviewed September 24, 2026.
• American Psychiatric Association. Intellectual Disability: DSM-5 fact sheet on the revised diagnosis, 2013. psychiatry.org, reviewed September 24, 2026.
• American Association on Intellectual and Developmental Disabilities. Defining criteria for intellectual disability, from Intellectual Disability: Definition, Diagnosis, Classification, and Systems of Supports, 12th edition. aaidd.org, reviewed September 24, 2026.
• Hassiotis A, Strydom A, Hall I, Ali A, Lawrence-Smith G, Meltzer H, Head J and Bebbington P. Psychiatric morbidity and social functioning among adults with borderline intelligence living in private households. Journal of Intellectual Disability Research, 2008, volume 52, issue 2, pages 95 to 106.
• Peltopuro M, Ahonen T, Kaartinen J, Seppälä H and Närhi V. Borderline intellectual functioning: A systematic literature review. Intellectual and Developmental Disabilities, 2014, volume 52, issue 6, pages 419 to 443.
• Kerr-Wilson C O, Mackay D F, Smith G C S and Pell J P. Meta-analysis of the association between preterm delivery and intelligence. Journal of Public Health, 2012, volume 34, issue 2, pages 209 to 216.
• Lanphear B P and colleagues. Low-level environmental lead exposure and children's intellectual function: An international pooled analysis. Environmental Health Perspectives, 2005, volume 113, issue 7, pages 894 to 899.
• Bath S C, Steer C D, Golding J, Emmett P and Rayman M P. Effect of inadequate iodine status in UK pregnant women on cognitive outcomes in their children: results from the Avon Longitudinal Study of Parents and Children (ALSPAC). The Lancet, 2013, volume 382, issue 9889, pages 331 to 337.
• Mani A, Mullainathan S, Shafir E and Zhao J. Poverty impedes cognitive function. Science, 2013, volume 341, issue 6149, pages 976 to 980.
• Gottfredson L S. Why g matters: The complexity of everyday life. Intelligence, 1997, volume 24, issue 1, pages 79 to 132.
• Flynn J R. Tethering the elephant: Capital cases, IQ, and the Flynn effect. Psychology, Public Policy, and Law, 2006, volume 12, issue 2, pages 170 to 189.
• Supreme Court of the United States. Atkins v. Virginia, 536 U.S. 304 (2002), opinion of the Court by Justice Stevens. Cornell Legal Information Institute.
• Supreme Court of the United States. Hall v. Florida, 572 U.S. 701 (2014), opinion of the Court by Justice Kennedy. Cornell Legal Information Institute.
• Supreme Court of the United States. Moore v. Texas, 581 U.S. 1 (2017), opinion of the Court by Justice Ginsburg, slip opinion. supremecourt.gov.
• United States Code. Title 10, section 520: Limitation on enlistment and induction of persons whose score on the Armed Forces Qualification Test is below a prescribed level. Cornell Legal Information Institute, reviewed September 24, 2026.
• Raiford S E and Courville T. Equivalence of Q-interactive and paper administration of cognitive tasks: WAIS-5. Q-interactive Technical Report 16, NCS Pearson, September 2024. pearsonassessments.com, reviewed September 24, 2026.

The interpretive framework is the Standards for Educational and Psychological Testing, jointly published by AERA, APA and NCME in 2014, and APA guidance on testing and assessment. Appropriate use requires evidence for the intended interpretation and decision, including its limitations.

14 Frequently Asked Questions