Definitions of intelligence: what experts share and where they split

0 The short answer

No single definition of intelligence has ever been agreed by researchers, but the definitions experts cite overlap on a core: the capacity to reason, solve problems, learn and adapt. Two documents anchor the field. The 1994 statement drafted by Linda Gottfredson, signed by 52 researchers, defines it as a very general mental capability, and the 1996 American Psychological Association task force report states that none of the many concepts "commands universal assent." An IQ test measures a defined, sampled part of that capability, and the Standards for Educational and Psychological Testing treat the definition as something a test developer must specify, not assume.

1 What Does the Word Intelligence Name, and Why Has It Resisted Definition?

The word intelligence names a construct, a concept that scientists use to organize observed differences in performance, and a construct can be defined only by saying what is included, what is excluded and how it will be measured. The Standards for Educational and Psychological Testing (AERA, APA and NCME, 2014) use the term in exactly that way: "the concept or characteristic that a test is designed to measure." Among its examples of constructs currently used in assessment, the Standards list general cognitive ability next to depression and self-esteem. They add that "rarely, if ever, is there a single possible meaning that can be attached to a test score," so that test developers and users must specify the construct interpretation they intend. That sentence frames the whole definitional problem. The reader is looking for the meaning of a word, while the measurement tradition asks which meaning a given score was built to support.

The early psychometricians felt this problem sharply. Charles Spearman, who in 1904 reported the general factor, wrote in 1927 that "intelligence" had become "a mere vocal sound, a word with so many meanings that it finally has none." Arthur Jensen went further in 1998, in a passage quoted in the same historical chapter, and recommended that psychologists drop the word from scientific vocabulary or use it only in quotation marks. Lewis Terman took the opposite practical line in 1916: he held that nobody should demand a complete definition before measuring, because, as Stern had pointed out, electrical currents were measured long before their nature was understood, and that an adequate definition could only be "based primarily on the symptoms empirically brought to light by the test method." John Wasserman, whose chapter in the handbook Contemporary Intellectual Assessment preserves these exchanges, observes that Boring and Terman both expected research to deliver a definition eventually, and asks how much longer the field must wait.

Three features make the word hard to pin down. It is an everyday word before it is a technical one, so every reader arrives with a private meaning. It covers behavior in the moment, stable differences between people and the biological capacity behind both, and a definition can aim at any of the three. And it carries value judgments, because what a society prizes as clever shapes what its experts include. The rest of this page treats those difficulties as evidence rather than as an embarrassment: how the definitions were collected, what they contain, who agreed to which statement, how an operational definition tries to sidestep the argument, and how far a test can serve each one. For the measurement side, the pages on what IQ measures and what IQ is cover the scale and the score, and this page stays with the concept.

2 What Did the 1921 and 1986 Symposia Find?

The two best known attempts to ask experts directly, a journal symposium in 1921 and an edited volume in 1986, produced many definitions and no consensus, though both were dominated by higher-level cognitive functions. The first was "Intelligence and Its Measurement: A Symposium," published in the Journal of Educational Psychology in 1921. Wasserman reports that the journal asked 17 leading investigators explicitly what they conceived intelligence to be. The journal's own record, as listed by Crossref, carries fourteen numbered contributions across three issues of volume 12, from Thorndike (I) and Terman (II) to Buckingham (XIV), with Freeman, Colvin, Pintner, Ruml, Pressey, Henmon, Peterson, Thurstone, Woodrow, Dearborn and Haggerty between them. The second was Robert Sternberg and Douglas Detterman's 1986 volume What Is Intelligence? Contemporary Viewpoints on Its Nature and Definition. The 1996 APA task force summarized the outcome in one sentence: when two dozen prominent theorists were asked to define intelligence, they gave two dozen somewhat different definitions. Wasserman counts 25 authorities asked to write essays, so the two descriptions differ by one; this page uses the task force's "two dozen."

Sternberg and Berg tabulated the attributes that recurred in each set of definitions, and Wasserman reports their percentages. The table below reproduces only the attributes he lists.

Two cautions apply. First, the shares are small counts. Our arithmetic shows that Wasserman's percentages match counts of 8, 4 and 3 definitions out of 14 for 1921 and of 12, 7, 6 and 5 out of 24 for 1986, so a difference of a few points is one or two definitions and is not evidence that the field changed its mind. Second, a blank cell means the attribute was not among those Wasserman lists, not that no definition mentioned it. What the table does support is a pattern: both panels put reasoning and thinking first, the earlier panel stressed adaptation and learning, and the later one added the cultural value of the behavior and executive control of one's own thinking. A similar contrast shows in the individual wordings in the next section.

Spearman had no patience for the exercise. Wasserman quotes his 1927 verdict on "repeated recourse to symposia" as a symptom of disagreement that had "reached its apogee." The symposia are still the best record of the problem, because each contributor is on the record with a sentence that can be checked.

3 A Dated Table of Definitions, and What Each One Stresses

Read in date order, the definitions keep returning to a handful of ideas, adaptation, learning, abstraction and achievement of goals, and they differ mainly on whether intelligence is one thing, many things, a kind of behavior or a measurement. The table collects 22 definitions with their dates. Wordings are short extracts. The rows from 1855 to 1986 follow John Wasserman's Table 1.2 (pages 37 and 38) in the fourth edition of Contemporary Intellectual Assessment, whose page references belong to the original works. The rows from 1994 onward come from the sources named after the table.

The sources behind the rows from 1994 onward are the Gottfredson 1997 reprint, the APA task force report, the Legg and Hutter collection, Sternberg's chapter on successful intelligence in the same handbook and Gignac and Szodorai's 2024 paper. Gardner's wording comes from the chapter by Jie-Qi Chen and Howard Gardner in that handbook, which quotes his 1999 definition.

Grouped by our own reading, not by any author's classification, the rows fall into five families. Adaptation definitions run from Spencer and Binet to Stern, Pintner and Anastasi. Learning and abstraction definitions include Woodrow, Terman and the 1994 and 1996 statements. Goal and success definitions run from Sternberg to Legg and Hutter and Gignac and Szodorai. Plurality definitions, from Haggerty and Detterman to Horn, deny that one process or one entity sits behind the word. A last family, Boring and Hunt, defines the term through the measurement itself. The families are not exclusive, and Wechsler's sentence touches adaptation, purpose and reason at once, which partly explains why it has lasted. For the question of whether one general capability sits behind the plural abilities, the page on the g factor holds the evidence.

4 What Do the Five Most Cited Definitions Share?

Five definitions do most of the work in modern discussion, and they agree on more than their wording suggests: each ties intelligence to a person or agent coping with problems or goals, and most name learning or adaptation. The five are Wechsler's, the 1994 statement's, the 1996 task force's, Legg and Hutter's and the 2024 proposal by Gignac and Szodorai.

Wechsler's phrase came first of the five and, as Wasserman suggests, owes its long life to the enduring popularity of the Wechsler scales that were built on it. His 1939 text, as Wasserman reproduces it, calls intelligence global because it characterizes the individual's behavior as a whole, and an aggregate because it is composed of abilities that, "though not entirely independent, are qualitatively differentiable." It adds that intelligence "is not identical with the mere sum of these abilities." That qualification keeps the definition from being a list.

The 1994 and 1996 definitions are in fact lists, and Gignac and Szodorai object to precisely that. They write that such definitions are "essentially a list of examples of subdimensions of intelligence, rather than representing an encompassing concept," and that an acceptable definition has to be both abstract enough and detailed enough to say what does and does not count. Their alternative defines human intelligence through three features: it is a maximal capacity and not typical behavior, the goal must be novel so that learned routines do not count, and the processes are perceptual and cognitive. As an example of why capacity differs from typical behavior, they note that the correlation between overall intelligence and typical intellectual engagement is only about 0.45. Legg and Hutter move the other way and delete the human reference entirely. The result is a spectrum from definitions that describe people to definitions that would apply to any system, and each end buys something: the first stays close to what tests sample, and the second allows comparison with machines.

5 What Did Gottfredson's 1994 Statement Say, and Who Signed It?

The 1994 statement "Mainstream Science on Intelligence" defined intelligence as a very general mental capability, and its signing record shows an endorsement by 52 researchers out of 100 who replied, not a poll of the whole field. Linda Gottfredson of the University of Delaware drafted it after the public controversy over Herrnstein and Murray's The Bell Curve, with a Wall Street Journal editor willing to publish it, and it first appeared in that newspaper on December 13, 1994. For the historical background of the debate over intelligence tests, see the page on the history of IQ testing. It was reprinted in the journal Intelligence in 1997 as "Mainstream Science on Intelligence: An Editorial with 52 Signatories, History, and Bibliography". Its first numbered conclusion is the definition: intelligence is "a very general mental capability that, among other things, involves the ability to reason, plan, solve problems, think abstractly, comprehend complex ideas, learn quickly and learn from experience." The same paragraph says what intelligence is not: "not merely book learning, a narrow academic skill, or test-taking smarts," but a broader and deeper capability for "catching on," "making sense" of things or "figuring out" what to do. The statement has 25 numbered conclusions in all. The first gives the definition, the next two address measurement, and the rest cover the bell curve, practical importance, group differences and policy; this page does not evaluate those.

Gottfredson's own history of the document supplies the figures behind the headline number. She built a list of experts from fellows of relevant divisions of the American Psychological Association, the editorial board of Intelligence, the tables of contents of books and journals on the subject, and suggestions from colleagues, and she invited only academics. In her account, 131 invitations were issued and 100 responses were obtained by a Friday deadline set in the same week the invitations went out. Of the 100 who replied, 52 signed and 48 declined. Of the 48, 7 said the statement did not represent the mainstream, 11 said they did not know enough to say, and 30 gave other reasons. She stresses that declining did not necessarily mean disagreement: 14 declined despite seeming to agree that the content was mainstream, some because they thought it should state more complexities and qualifications, or objected to a group statement or to a newspaper venue. Of 11 who explicitly disagreed with the content, 4 disputed the concept of general intelligence itself, calling it "not a useful concept." She also warns that no inferences can be drawn about who declined, because many worthy scholars were omitted or unavailable that week.

The figures put the title of this page in context. "52 signed, 48 declined" is a fact of the record, and the 48 are not a camp of opponents. The record supports a narrower reading: a group of 52 academics, invited by one author on a short deadline, agreed to a short statement of 25 conclusions, and the process left room for dissent on the status of general intelligence. Wasserman, who calls the statement and the APA report "relatively rare scientific consensus statements" on intelligence, adds that "there are many areas in which they appear to disagree." The next section looks at the second of the two.

6 What Did the 1996 APA Task Force Conclude About Defining Intelligence?

The 1996 American Psychological Association report did not choose a definition: it described what intelligence tests and theories try to explain, said that no concept of intelligence commands universal assent, and called the general factor the most widely accepted view while noting that some theorists regard it as misleading. The APA Board of Scientific Affairs created the task force, chaired by Ulric Neisser, in response to the same public controversy over The Bell Curve, and its report appeared as "Intelligence: Knowns and Unknowns" in American Psychologist. The task force had 11 authors and a stated charge "to prepare a dispassionate survey of the state of the art: to make clear what has been scientifically established, what is presently in dispute, and what is still unknown." That charge explains its caution. Where the 1994 statement wrote its first conclusion as a definition, the task force wrote a description of the phenomenon.

The description opens with people, not with a word: "Individuals differ from one another in their ability to understand complex ideas, to adapt effectively to the environment, to learn from experience, to engage in various forms of reasoning, to overcome obstacles by taking thought." The report goes on to say that none of the concepts of intelligence "commands universal assent," and it supports that with the 1986 symposium, in which two dozen prominent theorists gave two dozen somewhat different definitions. On structure, Wasserman quotes the task force as saying that "while the g-based factor hierarchy is the most widely accepted current view of the structure of abilities, some theorists regard it as misleading," and he reads this as hedging on g where the 1994 statement affirmed it.

For a reader who needs one citation, the task force report is the safer anchor for a definitional claim, because it was written to be neutral about what is in dispute. The 1994 statement is the better citation for the claim that a group of academics endorsed a particular definition, which is what its signing record documents. The two documents handle the general factor differently, and the page on whether IQ is real sets out the critiques that sit behind the task force's hedge.

7 Do Experts Agree on the Parts of Intelligence? The 1987 and 2020 Surveys

Surveys of experts show near unanimity on a core of abstract reasoning, problem solving and acquiring knowledge, and a real split on adaptation, creativity and motivation, which many respondents thought tests do not capture. The most informative survey on definitions is Mark Snyderman and Stanley Rothman's "Survey of Expert Opinion on Intelligence and Aptitude Testing" in American Psychologist in 1987. They mailed 1,020 questionnaires in September 1984 to a sample of social scientists and educators drawn from the membership lists of professional organizations, and 661 completed questionnaires came back, a 65 percent response. One question listed 13 behavioral descriptors, with room for write-ins, and asked respondents to check each one they believed to be an important element of intelligence; the response rate on that question was 93 percent. A follow-up asked which of the elements they had checked they believed the most commonly used intelligence tests do not adequately measure.

The table follows Table 2 of Snyderman and Rothman (1987). The authors describe three tiers: near unanimity of more than 96 percent among those who answered, a middle tier checked by 60 to 80 percent, and a tier checked by fewer than 25 percent.

The authors note that this question "essentially concerns construct validity." They read the result as substantial support for the validity of the commonly used tests, with two exceptions where a majority of those who valued the element said the tests miss it. One is adaptation to one's environment, which they link to the common criticism that tests are much better at measuring traits important to success in school than general life skills. The other is creativity, which they link to the poor correlation between intelligence tests and creativity tests. On the structure question, 58 percent favored some form of a general intelligence solution, 13 percent favored separate faculties, and 16 percent thought the data too ambiguous to favor either.

Heiner Rindermann, David Becker and Thomas Coyle ran a later survey of experts, "Survey of expert opinion on intelligence", published in Intelligence in 2020. According to its abstract, up to 102 experts answered the Internet-based Expert Questionnaire on Cognitive Ability, administered in 2013 and 2014, on IQ research, controversies and the media. That survey asks mostly about research questions and public debate, not about the elements of the definition, so it supports a different point: an expert panel is a sample of the people who agree to answer, and a mailed questionnaire from 1984 and an Internet questionnaire from 2013 and 2014 reach different panels. Their percentages should not be compared. Gignac and Szodorai add that expert definitions generally align with the implicit theories held by laypeople and by specialists in other fields, citing Sternberg's work from 1981 and 1985, which suggests that the shared core is not confined to specialists.

8 Is Defining Intelligence as What the Tests Test Circular?

Boring's 1923 sentence was offered as a deliberately narrow starting point, not as a claim that tests exhaust the concept, and its weakness is circularity, which later writers tried to repair by separating the construct from the score. Edwin Boring wrote in The New Republic, in the words Wasserman reproduces: "Intelligence as a measurable capacity must at the start be defined as the capacity to do well in an intelligence test. Intelligence is what the tests test." He continued that this "is a narrow definition, but it is the only point of departure for a rigorous discussion of the tests," and that no harm need result if we remember that measurable intelligence is simply what the tests test "until further scientific observation allows us to extend the definition." Wasserman calls it "frequently quoted but much reviled." The reviling usually ignores the clause about extension, which makes the definition provisional.

The circularity problem is real even so. If intelligence is defined as test performance, the claim that tests measure intelligence cannot be false, and nothing about a test can be evaluated against the construct. The Standards for Educational and Psychological Testing define validity as "the degree to which evidence and theory support the interpretations of test scores for proposed uses of tests." Evidence and theory need a construct that exists apart from the score, which a purely operational definition removes. Several repairs have been offered. Hans Eysenck, as listed in the Legg and Hutter collection, distinguished Intelligence A, the biological substrate of mental ability; Intelligence B, its manifestation in real life behavior; and Intelligence C, the level of performance on psychometric tests. Boring's sentence defines only the third. Earl Hunt treated the word as a shorthand for variation in competence on cognitive tasks, which makes it a label for a pattern of differences rather than a thing.

A structural answer has also been proposed. Wasserman reports the argument by Gottfredson and Saklofske that definitional issues are "now moot" because the empirical referents of the term can be distinguished empirically and related within a common conceptual structure, which they identify as the Cattell-Horn-Carroll model. In that model, as Joel Schneider and Kevin McGrew describe it, abilities are "theoretical entities inferred from the observed relations among specific abilities," organized into narrow and broad abilities with abbreviations such as Gf, Gc and Gwm. The CHC model therefore replaces one verbal definition with a map. Gignac and Szodorai accept the spirit and quote Humphreys: a scientist has "a duty to define concepts in a way compatible with measurement operations." On that view the definition and the test are written together and checked against each other, which is the opposite of Boring's shortcut. The page on reliability and validity explains how that checking is done.

9 Does Culture Change What Counts as Intelligence?

Several definitions build culture into the concept itself, and the evidence for that position is a set of specific arguments and one cited field study, not a general proof that tests are invalid in any particular group. James Pellegrino's 1986 wording says that individuals' responses to situations vary in quality and value "as judged by their culture." John Berry, in the same volume, called intelligence an end product of individual development in the cognitive domain that is "adaptive for the individual, permitting people to operate in their particular cultural and ecological contexts." In the Sternberg and Berg tabulation reported earlier, the attribute "that which is valued by culture" appears in 29 percent of the 1986 definitions, which is about 7 of 24 by our arithmetic. Howard Gardner's definition, quoted in the chapter by Jie-Qi Chen and Gardner, is a biopsychological potential "that can be activated in a cultural setting to solve problems or create products that are of value in a culture."

Robert Sternberg takes the position furthest. In his chapter in the same handbook, successful intelligence is the use of skills needed to attain success in life "however an individual defines it, within his or her sociocultural context." He reports that in research in rural Kenya, children who may score quite high on a test of practical knowledge, namely how to use natural herbal medicines against parasitic and other illnesses, may score quite poorly on tests of IQ and academic achievement, with an inverse relationship and correlations reaching negative .30. His interpretation is that time in school takes time away from learning the practical skills that families view as needed for success. That is one study as its author describes it, in one setting, and it does not show that IQ tests are invalid. It shows how a definition anchored in locally valued success produces a different list of abilities than one anchored in school performance. The page on types of intelligence covers Sternberg's and Gardner's models in full, and whether social and emotional abilities count as intelligence is the subject of the page on emotional intelligence and IQ.

The 1994 statement takes the other side on one point. Its third conclusion holds that, while there are different types of intelligence tests, they all measure the same intelligence, and that some require specific cultural knowledge such as vocabulary while others use shapes or designs and require knowledge of only simple, universal concepts. Whether that holds for a given test and group is an empirical question, and the Standards give it a name. Their glossary defines construct equivalence in part as "the degree to which a construct measured by a test in one cultural or linguistic group is comparable to the construct measured by the same test in a different cultural or linguistic group." The APA Guidelines for Psychological Assessment and Evaluation (2020) go further in Guideline 12, noting that assessment instruments "are culture specific because they are developed from a specific sociocultural context, most often Western," and that the validity and reliability of an instrument designed for one population are tied to its normative group. None of these sources settles whether the concept of intelligence is Western. They say that instruments are built in a context, that equivalence must be shown rather than assumed, and that some prominent definitions make the context part of the meaning. The pages on culture fair tests and whether IQ tests are biased deal with the evidence on equivalence directly.

10 How Do the Definitions Treat Machines and Other Agents?

Definitions written for artificial systems drop the human reference and keep the common core of goals, environments, adaptation and learning, and the psychologists who reviewed them argued that high task performance alone is not intelligence. Shane Legg and Marcus Hutter's 2007 collection, which describes itself as a survey of "70-odd" informal definitions, is the broadest single gathering. It has three numbered sections: definitions proposed by groups or organizations (18 entries), definitions from psychologists (35) and definitions from artificial intelligence researchers (18). That is 71 in all by our count of the numbered entries. The AI section reads differently from the psychologists' section. McCarthy calls intelligence "the computational part of the ability to achieve goals in the world," Minsky "the ability to solve hard problems," and Schank "getting better over time."

Legg and Hutter then scanned the whole collection for recurring features and found three. Intelligence is a property an individual agent has as it interacts with its environment or environments; it is related to the agent's ability to succeed with respect to some goal; and it depends on how able the agent is to adapt to different objectives and environments. Putting these together gives their own definition: "Intelligence measures an agent's ability to achieve goals in a wide range of environments." They state that features such as learning, adapting and understanding are implicit in it, because those capacities enable success across environments. They also write that it is difficult to argue that any one definition is objectively correct, only that some are more concise, precise and general than others. A formal version of the idea, which they call universal intelligence, appears in a separate paper, and the page on collective intelligence examines whether a group can be treated as a unit with an ability of its own.

Two later proposals push back on task performance. François Chollet's 2019 paper argues that skill is "heavily modulated by prior knowledge and experience," so that benchmarks of skill at specific tasks are an insufficient measure of intelligence, and proposes defining it as skill-acquisition efficiency, with a benchmark designed to compare machines and humans on a human-like form of general fluid intelligence. Gignac and Szodorai reach a similar place from psychometrics. They note that the common definition of artificial intelligence as the ability of machines to perform tasks that typically require human intelligence is circular, because it does not define human intelligence. Their parallel definition for machines uses maximal capacity for novel goals through computational processes, and they conclude that current evidence more greatly supports artificial achievement and expertise than artificial intelligence. They propose "AI metrics" as a discipline that would apply to artificial systems the reliability, validity and standardization requirements of psychometrics.

The practical consequence for a reader is narrow. An IQ is defined relative to a human reference group, as the page on how IQ scores are normed explains, so a number from a human test is not transferable to a machine without that reference group. The evidence on what language models have been scored on, and why those scores are not IQs, is on the pages about AI and IQ and ChatGPT's IQ. What the definitions add is a vocabulary: the Legg and Hutter definition applies to any agent and Gignac and Szodorai give parallel definitions for humans and machines, while Wechsler's and Gottfredson's apply to individuals, and a reader comparing humans to machines should state which definition licenses the comparison.

11 What Part of the Definitions Does an IQ Test Capture, and What Does It Leave Out?

A multi-domain IQ test samples the reasoning, knowledge, memory and speed parts of the definitions directly, samples learning, planning, adaptation in daily life, creativity and motivation little or not at all, and states its construct narrowly on purpose. The 1994 statement is explicit about the first half. Its second conclusion says that intelligence tests measure the capability it defines well, and that they "do not measure creativity, character, personality, or other important differences among individuals, nor are they intended to." The Snyderman and Rothman respondents drew the line in a slightly different place, as the table above shows: they did not think common tests measure adaptation or creativity adequately, and a large minority thought the same about the capacity to acquire knowledge.

The table below maps each element of the definitions to the closest broad ability in the Cattell-Horn-Carroll framework and says whether a six-index adult battery of the kind ACIS reports addresses it. The mapping is our reading, not Gottfredson's or anyone else's, and the ACIS column uses only the six indices the report names.

The CHC framework supplies the list of broad abilities. In Schneider and McGrew's chapter, Gf is a domain-general reasoning capacity, Gc, Gkn, Gw and Gq are acquired-knowledge capacities, and three further factors, Gwm, Gl and Gr, relate to memory, with learning efficiency (Gl) among them. The page on the CHC model and the page on cognitive domains describe each ability, and the page on fluid and crystallized intelligence separates the two that people confuse most often. The boundary with neighboring ideas has its own pages as well: intelligence versus knowledge, aptitude versus intelligence, critical thinking versus intelligence and intelligence versus wisdom.

The Standards name the risk exactly. Construct underrepresentation, in their words, "refers to the degree to which a test fails to capture important aspects of the construct." A test that fails to sample learning efficiency, planning or creativity does not fail as a measure of the narrower construct it states. It would fail only if it were presented as a measure of the whole definition. That is why the rest of this site reports an index profile and a Full Scale IQ as performance on named abilities, and why what IQ measures separates the general factor, the broad abilities and what lies outside both. The biological layer in Eysenck's first sense, the neural substrate of mental ability, has a separate evidence base, summarized on the page about intelligence and the brain.

12 How Should a Score Be Read When the Definition Is Contested?

A score should be read as a statement about performance on a stated set of tasks, compared with a stated reference group and carrying a stated margin of error, not as a verdict on intelligence in every sense, and the testing standards say the same. The Standards for Educational and Psychological Testing (AERA, APA and NCME, 2014) put the burden on the test developer. Standard 1.1 reads: "The test developer should set forth clearly how test scores are intended to be interpreted and consequently used. The population(s) for which a test is intended should be delimited clearly, and the construct or constructs that the test is intended to assess should be described clearly." Its comment adds that "no test permits interpretations that are valid for all purposes or in all situations." Applied to this page, a test that never says which abilities it samples, for whom, and for what use has not met the first requirement, however it defines intelligence on its home page.

The APA's Guidelines for Psychological Assessment and Evaluation, approved by the APA Council of Representatives in March 2020, say the same from the user's side. Guideline 6 asks psychologists to select tools that demonstrate "sufficient validity evidence for their uses, sufficient score reliability, and sound psychometric properties." Guideline 4 states that "validity is not a unitary property of the test instrument" and is evaluated within the settings and purposes of use. Guideline 7 observes that individual performance on tests "is only one piece of assessment." These guidelines address psychologists conducting assessments, and a person reading an online score is not one, but the reasoning carries over: a result is evidence to weigh with other evidence.

Five questions follow for any score, and each connects to a page on this site. Which construct does the test state, and which abilities does it sample? Which reference group and which norm date produced the number, as in the page on how IQ scores are normed? How large is the margin of error, as in the page on reliability and validity? Is the score reported as a percentile and a scaled number, as the page on IQ score versus percentile distinguishes? And does the use fit the instrument? The page on free versus validated IQ tests and the page on choosing a test and reading it apply these questions to concrete options.

ACIS is built to answer them in its own way, and its limits belong beside its design. It reports a Full Scale IQ and six index scores on the standard scale with a mean of 100 and a standard deviation of 15, with percentiles and a 95 percent confidence interval, from 20 subtests in six CHC domains. It is delivered online and unsupervised, it is not a clinical or diagnostic instrument, it is not for hiring, school accommodations or admission to high IQ societies, and it is in English only. Its technical manual is the place for the documentation. Its three forms, Quick, Optimized and Full Scale, cost 15, 30 and 50 dollars as one time payments when read on October 6, 2026, and prices can change. None of that makes it a measure of all of intelligence under any definition above. It is a measure of the six named domains, which is the claim the Standards ask a test to make and nothing wider.

13 Sources Behind This Page

Every definition, figure and quotation on this page comes from the sources below, opened on October 6, 2026, or from a source those documents quote and name. Early quotations from Boring (1923), Terman (1916), Spearman (1927), Jensen (1998), Binet, Stern and the 1921 and 1986 contributors are quoted as printed in Wasserman's chapter. The Eysenck, McCarthy, Minsky and Schank wordings are as listed by Legg and Hutter, the Gottfredson and Saklofske (2009) passage is as quoted by Wasserman, and the Humphreys (1984) passage is as quoted by Gignac and Szodorai. The counts of 8, 4, 3, 12, 7, 6, 5 and 71 definitions, and the figure of about 7 of 24, are our arithmetic on published numbers and are labeled as such where they appear. ACIS product facts and prices were read on the ACIS home page on October 6, 2026 and can change.

• Gottfredson L S. Mainstream science on intelligence: An editorial with 52 signatories, history, and bibliography. Intelligence, 1997, volume 24, issue 1, pages 13 to 23.
• Neisser U, Boodoo G, Bouchard T, Boykin A, Brody N, Ceci S, Halpern D, Loehlin J and 3 more. Intelligence: Knowns and unknowns. American Psychologist, 1996, volume 51, issue 2, pages 77 to 101.
• Snyderman M and Rothman S. Survey of expert opinion on intelligence and aptitude testing. American Psychologist, 1987, volume 42, issue 2, pages 137 to 144.
• Rindermann H, Becker D and Coyle T. Survey of expert opinion on intelligence: Intelligence research, experts' background, controversial issues, and the media. Intelligence, 2020, volume 78, article 101406.
• Intelligence and its measurement: A symposium (Thorndike, Terman, Freeman, Colvin, Pintner, Ruml, Pressey, Henmon, Peterson, Thurstone, Woodrow, Dearborn, Haggerty and Buckingham). Journal of Educational Psychology, 1921, volume 12, issues 3 to 5; the opening part is in issue 3, pages 123 to 147.
• Wasserman J D. A history of intelligence assessment: The unfinished tapestry (chapter 1, pages 3 to 55, Table 1.2 on pages 37 and 38). In Flanagan D P and McDonough E M (eds), Contemporary Intellectual Assessment: Theories, Tests, and Issues, 4th edition, Guilford Press, 2018.
• Schneider W J and McGrew K S. The Cattell-Horn-Carroll theory of cognitive abilities (chapter 3, pages 73 to 163). In Flanagan D P and McDonough E M (eds), Contemporary Intellectual Assessment, 4th edition, Guilford Press, 2018.
• Chen J-Q and Gardner H. Assessment from the perspective of multiple-intelligences theory: Principles, practices, and values (chapter 4, pages 164 to 173). In Flanagan D P and McDonough E M (eds), Contemporary Intellectual Assessment, 4th edition, Guilford Press, 2018.
• Sternberg R J. The triarchic theory of successful intelligence (chapter 5, pages 174 to 194). In Flanagan D P and McDonough E M (eds), Contemporary Intellectual Assessment, 4th edition, Guilford Press, 2018.
• Legg S and Hutter M. A collection of definitions of intelligence. Frontiers in Artificial Intelligence and Applications, 2007, volume 157, pages 17 to 24.
• Chollet F. On the measure of intelligence. arXiv, 2019, preprint 1911.01547.
• Gignac G E and Szodorai E T. Defining intelligence: Bridging the gap between human and artificial perspectives. Intelligence, 2024, volume 104, article 101832.
• American Educational Research Association, American Psychological Association and National Council on Measurement in Education. Standards for Educational and Psychological Testing. American Educational Research Association, 2014.
• American Psychological Association, APA Task Force on Psychological Assessment and Evaluation Guidelines. APA Guidelines for Psychological Assessment and Evaluation. Approved by the APA Council of Representatives, March 2020.

14 Frequently Asked Questions