Intelligence Testing: Uses and Limits

Theory · All theories
Evidence · 106 claims (35 for, 62 mixed, 9 against) · 56 studies (33 associational, 12 review, 6 theoretical, 3 design, 2 quant-synthesis), q1–q3 · 7 of 56 report an effect size · 105 claims rest on one study

Description

The first intelligence test, the Binet-Simon test, grew out of Alfred Binet's 1904 appointment as France's Minister of Public Education, where he sought a way to identify children with developmental disabilities so they could receive appropriate support — Binet believed "abnormality" could be remediated, and strongly supported special education on that basis. It was the first test to distinguish "normal" children from those with developmental disabilities, and its later revision became the Stanford-Binet, one of the most widely used intelligence tests (Zimmerman & Schunk, 2003). Binet's original purpose — identifying students who need support — sits in some tension with how intelligence-test scores are more often used today, as a general ranking or classification device (see Constraints below).

Most intelligence tests aim to measure "g," the general intelligence factor, and are judged by two standards: reliability (consistency over repeated administration) and validity (actually measuring intelligence rather than something else). Because what counts as an "intelligent" answer changes with age (a 3-year-old multiplying two two-digit numbers is precocious; a 25-year-old unable to do so is not average), interpreting a raw score requires standardization — testing a large, representative sample at each age level to establish population norms. Comparing a person's performance to these age norms yields their mental age; comparing mental age to chronological age yields the classic intelligence quotient (IQ) formula (mental age ÷ chronological age × 100), though most modern tests instead rank a person's raw score against same-age peers directly rather than computing this ratio literally.

Because population-wide intelligence-test performance drifts upward over time (see the Flynn effect [~M]), tests must be periodically re-standardized or their norms become stale and systematically overstate test-takers' relative intelligence. The Wechsler Adult Intelligence Scale (WAIS) is the most widely used adult intelligence test; its current version, the WAIS-IV, was standardized on 2,200 people aged 16-90, spans 15 tasks across four domains (verbal, perceptual, working memory, processing speed), and has very high reliability (>0.95) along with substantial correlations to real-world criteria — college grades, work performance, occupational level, and everyday functioning among people with intellectual disabilities. Downward and upward age-range adaptations exist for preschoolers (WPPSI-IV) and school-age children/adolescents (WISC-V).

Despite this psychometric strength, intelligence tests and the definitions of intelligence behind them have been heavily criticized since the 1970s on two related grounds. First, they have tended to favor Anglo-American, middle-class test-takers and to poorly capture non-academic forms of ability or talent. Second, what counts as "smart" is itself culturally variable in ways the tests don't accommodate: in dominant Western culture, speed is often equated with intelligence (the fastest correct answer is seen as smartest), whereas other cultures instead value a slower, more contemplative, thoroughly-considered answer as the mark of intelligence. A test built around one of these cultural assumptions will systematically undervalue test-takers operating under the other.

Implications

Context

Requirements

• Periodic re-standardization of any intelligence or ability test against a current, representative population sample
• Awareness of what a given test's format actually rewards (e.g., speed) versus what "intelligent" behavior looks like in a test-taker's own cultural context

Constraints

• Contemporary use for general ranking or classification departs from Binet's own founding purpose of identifying students who need additional support, not of sorting students overall
• Norms decay: a test standardized years or decades ago overstates the relative standing of current test-takers, since the population's average performance rises over time [~M]
• IQ scores do not reflect a person's ability to change — a test-blind spot with real consequences for how the resulting number gets used or interpreted downstream
• Tests calibrated to a speed-favoring or Anglo-American, middle-class conception of "smart" undervalue test-takers whose cultural background favors contemplative reasoning or values non-academic forms of ability

Target Learners

• Learners being formally assessed for eligibility decisions (giftedness, intellectual disability classification) where the underlying test's cultural assumptions and standardization currency directly affect the decision
• Learners from cultural backgrounds whose implicit definition of "smart" diverges from the speed-and-analytic-reasoning framing built into most standardized instruments

Target Learning Objectives

• Accurate, fair identification of a learner's actual capability, distinct from how well they happen to match a test's built-in cultural and format assumptions

Theory

Supporting

• Triarchic Theory of Intelligence [+M] — analytical-intelligence-focused standardized tests are exactly the instrument type Sternberg argues fails to capture creative or practical intelligence

Contradicting / Qualifying

• Multiple Intelligences Theory [~M] — if intelligence is genuinely multi-component, a single g-focused score cannot represent a learner's full range of ability

Claims

• Average scores on intelligence tests have risen substantially across generations worldwide [~M] — the basis for the requirement that tests be periodically re-standardized
• Teacher reports of noncognitive skills add predictive power beyond baseline IQ for most school outcomes [~W]
• Test scores measure accumulated knowledge at a moment in time, not innate student ability [+W]
• Children with ADHD perform worse than controls on WISC-IV Coding but not on Coding Copy, indicating deficits beyond graphomotor speed [~W]
• Curve estimation shows linear decline in processing speed after age 30, quadratic trajectories for category fluency and IQ, and no age change in letter fluency [~W]
• The set of executive-function and intelligence predictors of academic achievement differs across Grades 2, 4, and 6, with EF contributions appearing cumulative as students get older [~W]
• Full-scale IQ is genetically related but not redundant with common cognitive variance [~W]
• In predicting the change in achievement from 3rd to 4th grade, IQ is positively and study habits negatively associated with change, together predicting 12% [+W]
• IQ, short-term memory and study habits each correlate positively and significantly with academic achievement in both 3rd and 4th grade [+W]
• Full-scale IQ predicts reading comprehension more strongly than word reading, and working memory and processing speed remain significant predictors after IQ is controlled [+W]
• High performance in old age is associated with higher education, foreign language use and IQ, but not physical activity or personality [+W]
• Complexity of children's naturalistic memory reports predicts reading ability over and above IQ [+W]
• When previous academic achievement is controlled, intelligence predicts only later reading comprehension and age-10 mathematics, and WMC predicts nothing additionally [+W]
• Processing speed has a significant effect on both letter and category fluency in older adults, while IQ does not [~W]
• In latent-variable SEM models, intelligence — not working memory — is the significant direct predictor of academic outcomes in both childhood and early adulthood [+W]
• Specific processing speed measured at the start of Grade 7 predicts mathematics achievement on the high school entrance examination at the end of Grade 9 [+W]
• Specific processing speed generalises beyond mathematics, significantly predicting Chinese and English achievement in both Grade 7 and Grade 9 [+W]
• Fluid reasoning and self-control jointly explain far more variance in standardized test scores than in grades [+W]
• Males show a statistically significant but practically negligible advantage on WISC Full-Scale IQ (d = 0.09), which becomes non-significant with newer WISC versions [~W]
• Gender differences in fluid intelligence (gF) on the WISC are negligible, and newer fluid-reasoning subtests show no difference or favour females [~W]
• Instrument type moderated the mental rotation gender-difference effect size but not the visualisation effect size [~W]
• Cross-cultural studies show Eastern children can outperform Western ones on creativity measures, and children with little schooling can show high divergent thinking [+W]
• Published form-board age norms differ markedly across investigators, yielding discrepant diagnoses [~W]
• Group scales allow examining fifty or a hundred children in an hour, whereas individual Binet testing restricts use to special cases [~W]
• The editor predicts NRC group tests will make intelligence testing a routine school measure and open the way for adjusting school work to pupil needs [~W]
• The Binet-Simon scale measured intelligence development from ages seven to eleven with at least 96 per cent efficiency under the study's conditions [+W]
• The twelve-year tests of the Binet-Simon scale were unsatisfactory, with several individual tests too difficult for the children tested [-W]
• The Binet-Simon scale accurately graded fairly minute individual differences in mental ability within same-age groups, as validated against teacher and principal judgments [+W]
• Yerkes-Bridges coefficients of intelligence of 1.20 or more are attained by very different percentages of children at successive ages (34.5% at ages 4-5 down to 0% at ages 16-17 and 18-on), so they do not represent the same degree of brightness at all ages. [~W]
• Summing all test scores into one general-intelligence rating manufactures the appearance of a general factor [~W]
• Verbal test segments correlate more highly with the composite intelligence criterion than non-verbal segments [+W]
• In the Yerkes-Bridges alteration of the Binet Scale, equal point increments are assigned to digit-span steps whose growth intervals, as placed in the Binet Scale, are approximately 1, 3, 3, and 4 years. [~W]
• Exceptional Binet-Simon test outcomes were rare (4 per cent) and mostly attributable to error in adult judgments rather than the scale [+W]
• Correlation magnitude rises with the standard deviation of mental age of the group, so cross-group comparisons of the two traits are spurious [~W]
• Arbitrarily assigned point values in the Yerkes-Bridges scale yield implausible growth-rate results, such as intellectual ability increasing more than fifty per cent during the fourth year and only approximately fifteen per cent during the seventh. [-W]
• The low Alpha–school-standing correlation is difficult to explain and investigators report conflicting correlation results [~W]
• Pupils characteristically 'Bright' or 'Slow' in the early elementary grades remained in the same category in high school, in both graduate and non-graduate groups [~W]
• Easy opposites scores correlate with scholastic rank, faculty and graduates outscoring seniors and juniors [+W]
• Binet's norms were worked out for Paris working-class children and may need revision for other social classes and regions [~W]
• Any single measure type approximates the composite (.89, .87, .85) but has only limited value used alone [+W] — attached 2026-10-24 from Chassell et al. (1921), which proposed "Composite of all measures as criterion for evaluating classification measures".
• Combinations of two measure types correlate very highly with the composite (.96, .91, .98), with little difference between choices [~W] — attached 2026-10-24 from Chassell et al. (1921), which proposed "Composite of all measures as criterion for evaluating classification measures".
• Quality of work rather than time consumed proved the most reliable measure of intelligence across the battery [~W]
• Group tests combined with teachers' estimates correlate .77 with the Stanford Revision, questioning the necessity of individual examination [~W]
• The Binet-Simon and Yerkes Point Scale ratings of children run parallel and are practically identical up to age ten [~W]
• The test battery showed striking parallelism between growth of intelligence and age in unretarded children, with leaps at grade transitions [~W]
• The test yields an analytical diagnosis of classes and individuals, illustrated by a pupil profile identifying specific abilities needing improvement [+W]
• Most referred children in the sample rated below their chronological age on both scales [~W]
• Group tests correlate much lower with the Stanford Revision (.45 and .53) and only .37 with each other [~W]
• The Thorndike-McCall correlates about 0.50 with other single comprehension tests and 0.69 with a composite of group intelligence tests [~W]
• Practice effects on the Thorndike-McCall test over two or more hours of testing are very small [~W]
• Stanford-Binet mental age correlations with reading rise from grade III to grade VI [+W]
• Correlations between performance on different intellectual tasks fail to reach 1.00, indicating specialization of intelligence [+W] — attached 2026-10-24 from E. L. Thorndike et al. (1921), which proposed "Thorndike's definition of intellect as the power of good responses with respect to truth or fact, specialized by data and operation".
• Against Stanford Binet criteria for 300 grade 3–6 children, the mental index correlated .47, the educational index .52, crude score with mental age .80, mental with educational index .62, and Difference with I.Q. .21 [~W]
• The amount of advance credit earned on the Binet scale depends on the extent of testing, not the grade of intelligence of the pupils [~W]
• Scattering on the Binet scales does not support the claim that unequal development is the specific characteristic of the feeble-minded; normal pupils scatter most and imbeciles least [~W]
• The review endorses educational guidance by early intelligence testing and selection by tests at high school and college entrance [+W]
• Scattering is considerably greater on the Stanford-Binet than on the earlier Binet-Simon scales, so conclusions about scattering differ according to the scale used [~W]
• The author concludes that much remains to be done before group intelligence tests can be fully relied upon for trustworthy placing of individual pupils [~W]
• Army intelligence test rankings corresponded with officers' rankings of soldiers (average correlation 0.536 across twelve companies), suggesting the tests measured something beyond literacy [+W]
• The reviewer reports that Baldwin's repeated examinations show the IQ is fairly constant though subject to fluctuations, and that repeated examinations exhibit a definite practice effect [~W]
• Women delinquents scored somewhat below an army comparison group on the Stanford-Binet, but overlap was so great that other causes of delinquency must be sought [~W]
• Otis scores proved of practical value over a two-year trial for guiding junior high pupils' choice between a foreign language and manual work [+W]
• Rankings of the same 120 junior high pupils by the Otis and Illinois group intelligence tests disagreed by a median of 18 places, with a rank correlation of 0.642 [-W]
• The National Intelligence Tests, prepared by a committee under National Research Council auspices with $25,000 from the General Education Board, are probably the most reliable group tests [+W]
• Percentile-difference scales are crude and embody a fundamental fallacy because percentile units are not equal intervals [~W]
• The Haggerty Intelligence Examination Delta 1 for grades 1-3 correlates .68 with teachers' estimates [~W]
• Clement's five group intelligence tests intercorrelated only modestly, with the Indiana Mental Survey below 0.40 against all others [+W]
• Correlations between two forms of the same test (up to 0.94, averaging 0.83 in Colvin's data) far exceed between-test correlations, casting doubt that different 'general intelligence tests' measure the same element [~W]
• Teacher judgment ratings diverged from standardized test results for individual pupils [~W]
• Dividing the 120 pupils into four class-sections by one test would place 51.6 per cent in the wrong section according to the other test [~W]
• Thurstone-Brown agreement was closer than Otis-Illinois: median rank changes of 6.7 places (freshmen, r = 0.74) and 10.4 places (sophomores, r = 0.613), with 26 and 32.8 per cent wrongly classed [~W]
• Root argues that no current tests measure the real basis of superior intelligence, which he holds consists of ability to suspend judgment, freedom from suggestibility, and a critical attitude [-W]
• The more verbal the test material, the higher its correlation with school achievement, with a partial correlation of 0.69 for verbalness alone [+W]
• Verbalness outweighs test length as a factor determining correlations with achievement [~W]
• Adding the verbal group test to Stanford MA raises the multiple correlation with achievement from 0.54 to 0.605 [~W]
• Non-verbal test materials add little to predicting achievement beyond verbal tests and mental age [~W]
• Stanford-Binet mental age correlations with achievement rise with grade level, while verbal group test correlations stay about the same [~W]
• Poorest and best students differed most in speed-of-association tests but little in Marble Statue and Substitution tests [~W]
• Thorndike found correlations between Binet age and language tests higher than those between Binet age and a construction test, and therefore judged the Binet scale inadequate. [-W]
• The author judges the tests reveal class members' ability to do school work almost as accurately as months of teacher observation, supporting use for borderline promotion cases [+W]
• Taylor's Sheffield study suggests the Binet scale is broadly suitable for English children, though tests for ages 6 and 7 seem too easy and tests above 11 seem unsatisfactory. [~W]
• Terman and Knollin indicate the Stanford revision is more accurate than the Vineland revision in estimating dullness, borderlinity and feeble-mindedness. [~W]
• Travis, examining 128 University of Pennsylvania students, concluded the Binet prose-passage reproduction tests have no value for group application and should be eliminated for ages fifteen and over unless given under standard Binet conditions. [-W]
• Applying the Stanford revision to selected children in four California cities, Williams classified 32% feeble-minded, 20% borderline, 22% dull normal, and 26% normal or superior. [~W]
• Woolley found 25.5% of 18-year-old industrial workers feeble-minded on the point scale and advised changing the definition of feeble-mindedness so tested deficiency better matched common-sense opinion. [~W]
• Within a grade, younger pupils score higher on the Alertness Quotient than older pupils [~W]
• A test score represents a pupil's performance under given conditions, not the pupil's ability, and conditions vary with the examiner and the pupil. [~W]
• Selected academic groups differ in Alpha distributions, with Phi Beta Kappa electees concentrated in higher classes and probationers lower [+W]
• Alpha subtests differ in predicting grades, with correlations from .17 (Test 3) to .37 (Test 7) among seniors [~W]
• Correlations between Alpha and grades vary by class standing, lowest for graduates (r = .15) and highest for seniors (r = .38) [~W]
• Six independent applications of the Binet scale in widely separated countries show surprising agreement in the age ranking of the tests, suggesting many tests are correctly placed [+W] — attached 2026-10-24 from J. C. B (1912), which proposed "The Binet-Simon age-scale principle: arranging graded tests in a scale whose steps correspond to normal intelligence at each age from three to thirteen".
• Descoeudres' Geneva study of 24 pupils finds the Binet scale too easy at the bottom and too hard at the top, with bright pupils passing 57% versus dull pupils' 43% [-W] — attached 2026-10-24 from J. C. B (1912), which proposed "The Binet-Simon age-scale principle: arranging graded tests in a scale whose steps correspond to normal intelligence at each age from three to thirteen".
• Comparative data across six investigators lead the reviewer to recommend shifting fourteen individual Binet tests to different age levels, many back to their 1908 positions [+W]
• School and college continuance rise steeply with Army Alpha level among World War recruits [+W]
• In Goddard's Vineland examination of 1547 pupils, only about half of mentally over-age and under-age pupils show corresponding advancement or retardation in school grading [~W]
• Army mental tests demonstrated their value in selecting men for officers' training schools and other service lines requiring special ability [+W]
• The field's greatest need is an extensive experimental and critical study of the relative diagnostic value of available individual tests, followed by establishment of norms [+W]
• The Binet-Simon scale is a valuable diagnostic tool for classifying defective children when used by a trained examiner but dangerous in untrained hands [~W]
• The Binet scale can be understood and applied by the average teacher or school principal as readily as current pedagogical tests [+W]
• The Binet scale is heavily verbal — 43 of its 54 tests use words rather than acts — a criticism Ayres raises and the reviewer partially rebuts [~W]
• A mental appraisement of an individual is even more dependent upon the adequate training and quality of the examiner than upon the standardization of the tests used [~W]
• The large majority of freshmen scoring below 77.5 Alpha drop out before senior year [~W]
• Grouping pupils by mental age is not a reliable means of ensuring proper educational treatment [-W]
• Binet-Simon intelligence tests are applicable only to European and American conditions; other tests are needed for primitive peoples [-W]
• Relative to mental age, 68% of pupils do arithmetic work comparable to same-mental-age children, 59.8% in language, but only 13.5% in silent reading [~W]
• The reviewer argues mental tests are more needed for selecting superior children than for selecting inferior children [~W]

Related Principles

• Supporting Gifted and Talented Students — giftedness classification is conventionally IQ-threshold-based, inheriting all of this principle's measurement caveats
• Supporting Students with Intellectual Disabilities — intellectual disability classification likewise depends on IQ scores read alongside adaptive functioning, precisely because IQ alone is an incomplete and sometimes biased measure
• Criterion- and Norm-Referenced Testing — tests historically called "intelligence tests" are now more often called learning-ability, cognitive-ability, or scholastic-aptitude tests, a terminology shift reflecting the same controversy this page describes over whether "intelligence" implies fixed, inherited capacity rather than developed ability (Linn & Miller, 2005)
• Standardized Test Fairness and Bias — a parallel treatment of cultural and content bias in achievement and admissions tests specifically, distinct from but related to the IQ-test bias described here

Examples

• Follow the French two-stage placement plan: pedagogical examination first, then mental tests for equal mental retardation
• Research agenda for intelligence measurement: intercorrelations, IQ permanence, distribution form, and effects of coaching
• Use mental tests as supplementary evidence of pupils' abilities in dealing with the apparently bright and the apparently stupid

Key Sources

• Zimmerman, B. J., & Schunk, D. H. (Eds.). (2003). Educational psychology: A century of contributions. Erlbaum.
• Watkins, C. E., Campbell, V. L., Nieberding, R., & Hallmark, R. (1995). Contemporary practice of psychological assessment by clinical psychologists. Professional Psychology: Research and Practice, 26(1), 54-60.
• Flynn, J. R. (1999). Searching for justice: The discovery of IQ gains. American Psychologist, 54(1), 5-20. doi:10.1037/0003-066x.54.1.5
• Neisser, U. (Ed.). (1998). The rising curve: Long-term gains in IQ and related measures. American Psychological Association.
• Neisser, U. (1997). Rising scores on intelligence tests. American Scientist, 85(5), 440-447.
• Arduini-Van Hoose, N. (2020). Intelligence. In Educational psychology. Retrieved from https://edpsych.pressbooks.sunycreate.cloud. CC BY-NC-SA 4.0.

Page metadata

Field	Value
type	theory
id	intelligence-testing-uses-and-limits
title	Intelligence Testing: Uses and Limits
description	Standardized intelligence tests are among the most reliable instruments in psychology and predict a range of academic and occupational outcomes, but they must be periodically re-normed against the population, and they have long been criticized as culturally biased and blind to non-academic forms of ability.
status	review
canonical	True
generated	by claude/unspecified on 2026-08-29
trust	unverified — no one has confirmed this page's content

Frontmatter is the page's machine-readable record — see the Schema & Guide.