The Stroop test and IQ, and what the interference score measures

0 The short answer

The Stroop test is not an intelligence test, and the interference score it produces shares only a modest part of its variance with IQ. J. Ridley Stroop reported in 1935 that naming the ink colors of conflicting color words took 110.3 seconds against 63.3 seconds for plain colored squares, a difference of 47 seconds, and later work reads that slowdown as a mix of inhibition, goal maintenance and speed. The studies this page could open put the link with IQ at about negative .27 in one small adult sample, and a review of executive function research puts a common executive factor, which Stroop type tasks help define, at about 25 percent shared variance with g. A processing speed index captures the speed part of the task and none of the conflict part.

1 What Is the Stroop Test, and What Did Stroop Actually Find in 1935?

The Stroop test is a timed task in which a person must name the ink color of a printed word while ignoring what the word says, and Stroop's own 1935 data show that the conflict slows naming sharply but barely slows reading. The paradigm is old enough to be described freely. A word such as a color name is printed in an ink that spells a different color, and the instruction is to say the ink. The picture at the top of this page shows the format in its simplest form: a color word in a conflicting ink. Nobody needs a manual to feel the effect, because the written word is read before the instruction to ignore it can take hold.

The source is a single article, Stroop (1935), "Studies of interference in serial verbal reactions", in volume 18 of the Journal of Experimental Psychology, pages 643 to 662. The text of the article is public, and its three experiments are easy to summarize, as read on October 20, 2026.

In the first experiment, 70 college undergraduates read color names printed in conflicting inks and, separately, the same names printed in black. Reading the names in conflicting colors took an average of 2.3 seconds longer for 100 names, an increase of 5.6 percent, and Stroop described that increase as not reliable. The written word, in other words, was read almost as fast whatever ink it was printed in.

The second experiment reversed the task. One hundred students named the colors of plain squares and, in another condition, named the ink colors of conflicting color words. Naming squares averaged 63.3 seconds and naming the conflicting words averaged 110.3 seconds, an increase of 47 seconds that Stroop put at 74 percent. The asymmetry is the finding: ignoring a word while naming its ink is hard, while ignoring the ink while reading the word is easy.

The third experiment looked at practice. Over eight days, 32 undergraduates at the University of Arizona repeated the task. Practice decreased the interference from conflicting words but did not eliminate it. MacLeod's integrative review adds that Stroop himself produced a reverse effect in this experiment, in which the ink color interfered with reading the word, but only after considerable training and only transiently.

Three points about that 1935 study matter for the rest of this page. First, the quantity Stroop reported is a difference in seconds between two conditions, and every later interference score inherits that structure. Second, Stroop studied interference between verbal reactions, and nothing in the article treats the size of the difference as a measure of a person's intelligence. Third, the samples were college students, and the numbers describe group averages, so they say how much longer the conflict condition took for those groups and not what a given person should expect.

2 How Is the Stroop Test Scored, and What Is an Interference Score?

An interference score is a difference score: it compares performance on the conflict condition with what the person would be expected to do given their speed on the simpler conditions, so it is only as trustworthy as the two measurements it subtracts. The best single description of the clinical card version comes from Scarpina and Tagini's systematic review of the Stroop Color and Word Test in Frontiers in Psychology (2017), which this page read on October 20, 2026.

In that review the test has three conditions. In the word condition, color words printed in black ink are read aloud. In the color condition, color patches are named. In the color word condition, color words printed in an incongruent ink must be answered by naming the ink, which means performing a less automated task and inhibiting the reading of the word. The first two conditions set a baseline of how fast the person reads and names, and the third adds the conflict.

How the three are combined is not settled. The most common method in the review is Golden's (1978), in which the interference score equals the color word result minus a predicted value computed from the word and color results, the product of the two divided by their sum. A negative value under that formula is read as a weak ability to inhibit interference. The review also lists other formulas, such as a time based index that subtracts the average of the word and color times from the color word time. Its authors argue that none of the reviewed methods fully captures the Stroop effect, and they recommend counting correct answers within a fixed time in all conditions together with a global index based on the color word condition.

The computerized versions that researchers use are different again. In Hedge, Powell and Sumner's study of the reliability of classic cognitive tasks, participants responded by key press to the color of a single centrally presented word, and the Stroop score was the reaction time cost, the difference between incongruent and congruent conditions. Checa and Fernández-Berrocal used a numerical version of the task in which the interference index was incongruent minus congruent reaction time, with a separate index for errors. A paper that reports a "Stroop correlation" may therefore be describing a printed card scored by a formula, a key press task scored in milliseconds, or a numbers variant, and these are not interchangeable.

Raw interference is also not a standard score. Van der Elst and colleagues administered the Stroop Color-Word Test to 1,856 cognitively screened, healthy Dutch speaking participants aged 24 to 81 and built normative data stratified by age, gender and education, because those variables changed the scores. A score of 47 seconds of added time means different things at different ages, which is the same reason an IQ is expressed against peers of the same age. The page on how IQ scores are normed explains that logic for intelligence tests, and the page on the standard deviation of 15 explains the unit that IQ and index scores share.

3 Is the Stroop Test an Intelligence Test?

No: the Stroop belongs to the family of executive function tasks, and the batteries that publishers sell for intelligence do not contain it. Two product pages, read on October 20, 2026, show the division of labor.

Pearson's page for the Delis-Kaplan Executive Function System lists a Color-Word Interference Test as one of nine tests, beside a Trail Making Test, a Verbal Fluency Test, a Design Fluency Test, a Sorting Test, a Twenty Questions Test, a Word Context Test, a Tower Test and a Proverb Test. It gives a publication year of 2001, an age range of 8 through 89 years and a qualification level of C, and says the system assesses key components of executive functions believed to be mediated primarily by the frontal lobe. The interference task is housed in an executive battery, administered by a professional with the credentials the publisher requires.

Pearson's overview brochure for the WAIS-5, the 2024 Wechsler scale for ages 16 through 90 years and 11 months, shows what an intelligence battery contains. Its primary battery has 10 subtests across five domains: verbal comprehension, visual spatial, fluid reasoning, working memory and processing speed. Counting the ancillary subtests, it lists 20 tasks in all, among them Similarities, Vocabulary, Block Design, Matrix Reasoning, Digit Sequencing, Coding and Symbol Search. None of the 20 is a color word interference task. The page on the WAIS-5 subtests describes them one by one.

The same split holds for the instrument this page belongs to. ACIS reports a Full Scale IQ and six primary indices from 20 subtests, and none of the 20 is a color word interference task. The six domains are verbal comprehension, fluid reasoning, quantitative reasoning, visual spatial, working memory and processing speed, and none is an inhibition domain. The page on what IQ measures and the page on the CHC model explain why batteries are organized around broad abilities such as Gc, Gf, Gv, Gwm and Gs, and the page on what each type of subtest measures places individual tasks inside that scheme.

None of this means the Stroop is unrelated to intelligence. It means the question is empirical, and that the field has spent decades on it. MacLeod (1991) wrote that the articles on the Stroop effect numbered in excess of 700 and that at least 300 of them used Stroop interference as a measure of attention in studying individual differences, drug effects and similar topics. A task used that heavily as a general index of attention invites the question of whether it tracks general ability. The sections below review what the studies this page could open found, and they are stated narrowly because the evidence is thinner than the task's fame suggests.

4 What Does the Interference Score Measure?

The interference score is a mixture: researchers have attributed it to inhibition, to keeping the task goal active, to the relative speed of reading and naming, and to the person's baseline speed, and no single account has survived intact. The label in popular use, "the inhibition test," is one reading among several.

Scarpina and Tagini describe the test as assessing the ability to inhibit cognitive interference, and they argue that performance relies on a specific executive, frontal domain. They also note that earlier work has used the test to measure attention, processing speed, cognitive flexibility and working memory, and that the reviewed studies vary in whether they interpret scores as executive, attentional or a sign of general cognitive efficiency. They add two points that bear directly on the speed question: scoring only reading time can hide inhibition difficulties, and general slowing, which can have many causes, affects performance across all conditions and not only the conflict one.

The deepest review of the mechanism remains MacLeod's (1991). It organized some 400 studies, isolated a set of 18 reliable empirical findings that any theory of the Stroop effect must capture, and judged the two leading theories, relative speed of processing and automaticity of reading, to be wanting. It concluded that theories placing the weight on parallel processing of the relevant and irrelevant dimensions were more likely to succeed than theories that look for a single bottleneck in attention. That is a verdict on mechanism, and it matters for intelligence research because "automatic reading beats a controlled response" turns out to be too simple a story to hang an ability on.

A second line of work ties the effect to working memory. Kane and Engle (2003) found in five experiments that working memory capacity predicted Stroop performance. When a version of the task contained many congruent trials, such as the word RED printed in red, people with lower working memory capacity made more errors on the rare incongruent trials, because they were less able to keep the "ignore the word" goal available. In versions with no or few congruent trials, working memory predicted the size of the reaction time interference instead. The authors concluded that interference is jointly determined by goal maintenance and competition resolution, and that the weight of each depends on working memory capacity and on the task set the context induces. The practical lesson is that the same task can reward different abilities depending on how many trials are congruent, so two "Stroop scores" from different versions are not the same measurement. The page on working memory tests and the page on the Digit Span subtest describe how working memory itself is measured in intelligence batteries.

The factor structure of inhibition is also contested. A later review by Friedman and Miyake (2017) maps the inhibiting tasks of the original latent variable work to antisaccade, stop-signal and Stroop tasks, and it attributes the low correlations among executive tasks in general to low reliability, different strategy use and task impurity, which it calls an unavoidable quality of executive tasks.

The table sets out the accounts and what each can and cannot explain.

None of these accounts is a statement about intelligence. Each is a statement about why a single timed conflict costs what it costs, and a task whose score has four candidate sources will not correlate cleanly with anything.

5 How Strongly Does Stroop Interference Correlate With IQ?

The studies this page could open put the link between Stroop interference and IQ at small to moderate and scattered across designs, and none of them is a meta-analysis or a large adult sample tested on a full Wechsler battery. This page searched for a reliable pooled estimate of the correlation between Stroop interference and general ability and did not find one. What follows is the set of studies that could be opened, with the design behind each.

Several features of the table deserve to be read before any number is quoted.

The cleanest single coefficient is Checa and Fernández-Berrocal's negative .27 between reaction time interference and the Kaufman Brief Intelligence Test total. The sign is negative because interference is a time cost, so larger costs went with lower scores. Squaring the coefficient gives about 7 percent of shared variance, which is our arithmetic and not a figure the authors state. The sample was 92 mostly female university students on a numerical version of the task, the authors themselves describe the sample as small and limiting for generalization, and the error based index showed no relationship at all. Vocabulary, a crystallized measure, correlated about as strongly as Matrices, a fluid one, so the link did not favor the fluid measure.

Arffa's study concerns children and adolescents, including gifted ones, and this page does not carry it over to adults. The abstract reports that Full Scale IQ was significantly related to the color word condition and that gifted children outperformed other children on the executive but not the non-executive tests. Those are group comparisons in youth and not an adult effect size.

The latent variable rows state numbers for a different object. A common executive factor is what the tasks share, and the Stroop is one indicator among several. The next section takes that distinction up.

Two cautions complete the picture. First, the studies above differ in task version, intelligence measure, age group and analysis, which is what scattered means in practice. Second, correlation with IQ is a statement about people in a sample, and it does not say that a particular person with a large interference score has a lower IQ, or that a lower IQ explains a particular score. In the language of the g factor, the question is how much the task draws on general ability, which the page on the g factor explains. Whether that loading changes across ability levels is a separate and disputed question treated in the sibling page on Spearman's law of diminishing returns, and none of the studies above was designed to test it.

6 What Do Latent Variable Studies Say About Inhibition, Executive Function and Intelligence?

Latent variable work finds that the executive functions are related but separable, that what the inhibiting tasks share with other executive tasks relates to intelligence while no separate inhibiting factor could be extracted, and that the size of the link depends on the model. A latent variable analysis uses several tasks per function so that the shared part of the tasks, and not the idiosyncrasies of any one task, is what gets correlated with an outcome.

The framework behind these studies distinguishes three executive functions: shifting between mental sets, updating information in working memory and inhibiting prepotent responses. The Friedman and Miyake review places the Stroop among the tasks that define the third, alongside antisaccade and stop-signal tasks.

Intelligence came into the picture in Friedman and colleagues (2006). They related fluid intelligence, crystallized intelligence and WAIS IQ to three separable executive functions and reported that updating was highly correlated with the intelligence measures but inhibiting and shifting were not. In structural equation models that controlled for the correlations among the three functions, updating stayed strongly related to intelligence while the relations of inhibiting and shifting to intelligence were small and not significant. The authors concluded that current intelligence measures do not equally assess a wide range of executive control abilities.

The later review by Friedman and Miyake (2017) reframes the same result. It reports that in the earlier analysis only updating was significantly related to intelligence, and it interprets the small relations of inhibiting and shifting as overlap with updating, that is, with what the functions share. In a bifactor model run on more data from the same twin sample, intelligence correlated .51 with a common executive factor, .49 with an updating specific factor and negative .24 with a shifting specific factor. The authors state that the common factor is not g and shares only about 25 percent of its variance with g. They also report that no inhibiting specific factor could be extracted, because the common factor explained all the correlations among the inhibiting tasks, which they describe as isomorphic with the common factor.

Put together, these papers say something precise about the Stroop and IQ. What carries the relationship to intelligence in the latent models is the shared executive component, and working memory updating contributes about as much. A Stroop interference score contains that component along with task specific variance, measurement error and speed, so in our reading its observed correlation with IQ is a diluted version of the latent one. A factor correlation of .51 and an observed correlation of negative .27 are therefore not in conflict, because they describe different objects. The same review notes that the common factor was separable from speed, a point the speed section returns to.

A skeptical result completes the set. Rey-Mermet, Gade and Oberauer (2018) measured inhibition in young and old adults with 11 established laboratory tasks, among them a color Stroop and a number Stroop. The measures from individual tasks had good reliabilities but correlated only weakly with each other. A two factor model emerged, one for inhibition of prepotent responses and one for resistance to distracter interference, but it had low explanatory power, and the authors argued that the findings call into question inhibition as a psychometric construct. If the tasks hardly agree with each other, a single Stroop score cannot be taken as an index of a trait that all of them are supposed to share, and it certainly cannot be assumed to index intelligence.

7 How Reliable Is an Interference Score, and Why Does That Limit Its Link With IQ?

A difference score is usually less reliable than the two scores it subtracts, and the Stroop is no exception: the best test retest estimates this page could open are around .6, which caps how closely any interference score can track another variable. Reliability here means consistency of the score across occasions, and the page on reliability and validity explains how it is judged for intelligence tests.

Hedge, Powell and Sumner (2018) tested the reliability of seven classic cognitive tasks, among them the Stroop, in three studies of university students. Participants completed the tasks in two sessions three weeks apart. The Stroop appeared in the first two studies, where 47 and 56 participants entered the Stroop analyses after exclusions for low accuracy. Across the seven tasks, reliabilities ranged from 0 to .82. The Stroop reaction time cost had intraclass correlations of .6 and .66 in the first two studies, and it was one of only two response control indices that exceeded the authors' standard of good or substantial reliability, set at .6. The other was commission errors on a go/no-go task, at .76 in both studies. Their explanation is the heart of what they call the reliability paradox: tasks become well established because the effect is robust, which happens when people differ little from each other, and low variation between people makes the task a weak tool for ranking people. The authors cite a multi-laboratory project in which the Stroop effect replicated in every attempt, and on their argument that very robustness is the reason it ranks individuals poorly.

The same paper shows why subtraction costs reliability. The reaction time costs were generally lower in reliability than their components, and the total variance in the difference scores was reduced, often by a factor of 3 or 4, relative to the components. Our reading of that result is that subtracting a congruent time from an incongruent time removes much of what differs stably between people, such as overall speed of responding, and leaves a smaller remainder.

MacLeod's review records that opinions about the reliability of the card version were mixed in the older literature. Some authors were pessimistic and others, including Jensen in 1965, judged it good, with Jensen concluding that with multiple administrations the Stroop was probably more reliable than any other psychometric test. Those statements concern a different scoring and design from the computer tasks Hedge and colleagues tested, and they do not contradict the finding that a single reaction time difference is a modest measure.

Reliability matters for the IQ question through a standard psychometric rule: an observed correlation between two measures cannot exceed what their reliabilities allow, and it is reduced from the true correlation by roughly the square root of the product of the two reliabilities. As an illustration with assumed numbers, and not a result from any study above, a Stroop score with a reliability of .6 correlated with an intelligence measure of reliability .9 would show an observed correlation of about .29 even if the true correlation between the underlying constructs were .40. That is our arithmetic. It shows why a number in the .2 to .3 range can coexist with a stronger underlying relationship, and also why it is not safe to read an individual's interference score as stable. The page on timed tasks and reliability covers a related issue, that timing complicates reliability for any speeded measure.

8 Is the Stroop Effect Partly a Speed Effect?

Yes, in two senses: the conditions that anchor the score are speed tasks, and much of what looks like a group or age difference in interference shrinks once general slowing is taken into account. The word and color conditions of the card version are rapid reading and naming, and the formulas in Scarpina and Tagini's review exist precisely to adjust the color word time for how fast the person is on those two. In our reading the adjustment is imperfect, because the conflict condition is slower for everyone and a slower person loses more seconds in absolute terms even when the proportional cost is the same.

The clearest demonstration is the aging meta-analysis by Verhaeghen and De Meersman (1998). It pooled 20 studies from 15 articles that compared younger and older adults on the Stroop interference effect and found no significant difference between the age groups in the standardized interference effect, with a mean standardized difference of 2.04 for younger adults and 2.17 for older adults. A single regression line with a slowing factor of 1.9 described the data well, with an R squared of .83, and no age by condition interaction appeared. The authors concluded that the apparent age sensitivity of the Stroop interference effect appears to be merely an artifact of general slowing. In our plain reading, older adults were slower on every condition by about the same factor, and the conflict did not cost them disproportionately once that factor was applied.

That result is the pattern a general slowing account predicts, and the page on processing speed and the page on how IQ changes with age place that account inside intelligence testing.

The relationship between speed and executive control in younger people is not trivial either. Karhadkar and colleagues (2026) studied 797 twins in the Colorado Longitudinal Twin Study, who completed 3 processing speed tasks and 6 to 9 executive function tasks in late adolescence (mean age 17.26) and again in established adulthood (mean age 29.34). Processing speed at age 16 and processing speed at age 29 correlated .90, and processing speed correlated .70 with a common executive factor at age 17. The common factor and an updating specific factor in late adolescence predicted new variance in later processing speed (correlations of .13 to .16), but earlier speed did not predict new variance in the common factor. The authors concluded that speed and executive functions are highly correlated yet distinct constructs. That study used a common factor from many tasks and not the Stroop alone, so its .70 does not transfer to a single interference score, but it makes the direction of the point clear: speed and executive control overlap heavily.

The overlap is not complete. The Friedman and Miyake review reports that a perceptual speed latent variable did not absorb the common executive factor, which was separable from speed. And Van der Elst and colleagues' norming study of 1,856 adults found that the speed dependent scores, meaning the time to complete a subtest, were profoundly affected by age, gender and education, much more than the accuracy measures were.

Taken together, the evidence supports a modest, careful sentence: a Stroop interference score contains a speed component that overlaps with what a processing speed index measures, and it contains something else that speed does not fully explain. How large each part is for a given person cannot be read from the score.

9 What Does a Processing Speed Index Capture of the Stroop, and What Does It Leave Out?

A processing speed index captures the baseline speed that the first two Stroop conditions measure and none of the conflict that the third adds, which is why a speed index and an interference score are related but not interchangeable. The WAIS-5 and ACIS both build speed from simple, high volume tasks with no incongruent condition.

According to the Pearson WAIS-5 brochure, read on October 20, 2026, the primary Processing Speed index uses Coding and Symbol Search, and the full list of subtests places a third, Naming Speed Quantity, under processing speed. The brochure describes the broader processing speed coverage as verbal and visual. In our reading, the naming task is the nearest relative of the Stroop word and color conditions, because rapid naming is also a baseline speed task, but the brochure's list contains no conflict version of it.

ACIS reports a Processing Speed Index from two subtests, Symbol Search and Coding. In the Coding subtest the person matches symbols to letters as quickly as possible, and in Symbol Search the person scans symbols quickly and accurately, as the subtest pages describe: see the pages on Coding and Symbol Search. Neither contains a conflict condition. The report gives the index on the standard scale with a percentile and a 95 percent confidence interval, as described on the page on a normed processing speed test. ACIS does not offer a color word interference task, and this page does not present the Stroop as an ACIS measure.

What does this mean for someone who wants to know "how I would do on a Stroop"? Three things follow from the evidence reviewed above.

First, a high speed index would be expected to go with fast performance on the word and color baselines, because those are the same kind of work, and a low one with slower baselines. Because the interference score is computed relative to those baselines, a person with fast baselines still has to resolve the conflict, and that cost is not in the index. Second, the part of the Stroop that goes beyond speed, which Friedman and Miyake describe as separable from speed in a latent variable analysis, is not captured by any single ACIS subtest, and in our reading a profile that shows speed and working memory side by side is the nearest the battery comes to it. The page on index score differences explains how often a gap between two indices is unremarkable. Third, the limits of the instrument apply: ACIS is an online, unsupervised, English only assessment, it is not a clinical or diagnostic instrument, and speeded subtests in particular are sensitive to the device and the input method, which the page on data quality in online cognitive testing discusses.

On prices, as read on October 20, 2026 and subject to change, the Full Scale form covers all 20 subtests in six domains, so it includes the speed index, for a one time payment of 50 dollars with no subscription, and a free trial needs no card. The Quick form costs 15 dollars for 6 subtests in 3 domains, and the Optimized form costs 30 dollars for 13 subtests in 5 domains.

10 What Else Moves a Stroop Score: Age, Education, Practice and Reading?

Age, education, practice and reading skill all move Stroop scores, so a raw interference value is uninterpretable without norms for the version used and a reference group like the person tested. The norming study by Van der Elst and colleagues (2006) is the most direct evidence for adults. In 1,856 cognitively screened, healthy Dutch speaking participants aged 24 to 81, the speed dependent scores were profoundly affected by the demographic variables, and an interaction between age and low level of education appeared for the Stroop Interference score and for errors on the color word subtest. The authors read that as consistent with the idea that executive function, as measured by the Stroop, declines with age and that the decline is more pronounced in people with a low level of education. That is an interpretation within the paper, and a norming sample compares people of different ages, so it describes differences among people and not change within people.

The developmental picture reviewed by MacLeod (1991) is a U shape. Comalli and colleagues, testing people aged 7 to 80, observed the greatest interference in young children, a decline into adulthood and an increase again with advanced age, and MacLeod notes that later studies confirmed the decrease through adulthood and the increase after 60. The aging meta-analysis above qualifies the older end of that curve by showing that the standardized effect did not differ between younger and older adults once general slowing was modeled.

Reading skill matters at the young end of the curve. MacLeod describes an unpublished dissertation by Corbitt (1978) in which intermediate and good readers read words faster than they named colors and showed a Stroop effect without a reverse effect, while poor readers named colors faster than they could read words and showed the opposite pattern. That pattern shows that the size, and even the direction, of the effect can depend on how fast a person reads relative to how fast the person names, although MacLeod judged the relative speed account insufficient as a full theory.

Practice matters at any age. Stroop's third experiment, described above, found that eight days of practice reduced the interference without eliminating it, so a person who has taken a Stroop task before is not in the same position as a naive one. The page on how to prepare for an IQ test covers the broader question of retesting and familiarity on intelligence measures.

Education deserves one further comment because it sits on both sides of the IQ question. Education is related to both IQ and, in the norming data above, to Stroop scores, so part of any observed association between the two could run through schooling. The studies read for this page do not separate that path, and the page on average IQ by education level discusses the same entanglement for intelligence scores.

11 How Should a Stroop Result Sit Next to an IQ Profile?

A Stroop result is best read as a separate observation about speed and control under conflict, never converted into an IQ and never subtracted from one. This page found no peer reviewed conversion from interference to an IQ score, and the evidence above explains why none is expected: the shared variance is small, the score is only modestly reliable, and the relationship varies by version.

Where do Stroop scores come from in practice? In clinical work they come from executive function batteries given by qualified professionals, as the Pearson page for the Delis-Kaplan system shows with its qualification level of C. The page on neuropsychological testing for adults describes that route and its cost, and the page on who can administer an IQ test explains the qualification levels. Online versions of the task return a reaction time difference, and when the test is taken without supervision the timing depends on the device, the browser and the input method. This page did not survey online Stroop tools and makes no claim about any of them, and it asks one question of any of them: which published norms, in which version, for which age and education group, stand behind the number?

If a reader holds both an IQ profile and an interference score, the roles divide this way. The IQ and its indices describe the level and pattern of broad abilities relative to age peers, with a confidence interval around each. The interference score describes the extra time one conflict cost in one sitting, with the reliability limits described above. The Friedman and Miyake review states that the common executive factor is not g and shares about 25 percent of its variance with it, so the two numbers are neighbors and not substitutes.

A gap between them is also not a finding about the person. Group statistics are not individual diagnoses, and no study reviewed here supports reading a particular pair of scores as evidence of a condition. The page on ADHD and IQ separates executive function from intelligence for readers who arrive with that question, and anyone who is worried about attention or control should talk with a qualified clinician, who can use norms and history that a web page cannot.

One more practical point concerns variability. The test retest estimates of about .6 for a reaction time cost mean that a person repeating the task three weeks later could land noticeably elsewhere without any real change. A single Stroop result is a snapshot, and an IQ profile with a stated confidence interval is a more stable summary of how a person compares with others.

12 What Do the Testing Standards Say About Reading an Interference Score?

The Standards for Educational and Psychological Testing (AERA, APA and NCME, 2014) ask that every interpreted score rest on a clear construct, evidence for the processes claimed, and reliability evidence for that specific score, and an interference score meets that bar only for particular versions and uses. The Standards are published jointly by the American Educational Research Association, the American Psychological Association and the National Council on Measurement in Education, and four of them bear on this topic directly.

Standard 1.1 says the test developer should set out clearly how scores are to be interpreted and for which population, and that the construct the test is intended to assess should be described clearly. Its comment points out that it is incorrect to use the unqualified phrase "the validity of the test," because each recommended interpretation for a given use requires validation. "The Stroop measures intelligence" and "the Stroop measures inhibition" are both unqualified statements of this kind. The evidence above supports narrower ones, for example that a specific version, scored a specific way, correlates modestly with a specific intelligence measure in a specific sample.

Standard 1.12 asks that when a score interpretation depends on premises about the psychological processes of test takers, theoretical or empirical evidence for those premises be provided. Calling the interference score a measure of inhibition is a premise about processes, and the reviews cited in this page show that it is contested: the individual inhibition tasks correlate only weakly with each other, and one large study questioned inhibition as a psychometric construct.

Standard 1.14 asks that when interpretation of subscores, score differences or profiles is suggested, the rationale and evidence for it be given. An interference score is a score difference by construction, and the standard's comment adds that the distinctiveness and reliability of separate scores should be demonstrated. Standard 2.3 asks for estimates of reliability and precision for each total score, subscore or combination of scores that is to be interpreted, and its comment warns that reliability for total scores does not guarantee it for subscores. The reliability evidence in this page, about .6 for a reaction time cost across three weeks, is the kind of estimate those standards require to be reported and weighed.

The relevant APA guidance points the same way. The APA Ethical Principles of Psychologists and Code of Conduct, in the 2017 text read in a public copy on October 20, 2026, ask in Standard 9.02(a) that psychologists use assessment techniques in a manner and for purposes appropriate in light of the evidence on their usefulness and proper application. Standard 9.02(b) asks for instruments whose validity and reliability have been established for use with members of the population tested, and, where they have not been, a description of the strengths and limitations of the results. Standard 9.06 asks psychologists, when interpreting results, to take into account the purpose of the assessment and the test factors, test taking abilities and personal, linguistic and cultural characteristics of the person, and to indicate any significant limitations of their interpretations.

Applied to the reader: an interference score taken from a particular version, with norms matched to age and education, is a defensible observation when a trained person interprets it with its limits stated. An unnormed score from an unsupervised online task is a timing difference, and its meaning for intelligence is, on the evidence reviewed here, small and uncertain. The same standard applies to ACIS. It is an online, unsupervised, English only assessment, it is not a clinical or diagnostic instrument, and it is not meant for hiring decisions, school accommodations or admission to high IQ societies. Its documentation is on the technical manual page and is not quoted here, and the page on reliability and validity explains the concepts a reader needs to judge any such documentation.

13 Sources Behind This Page

Every figure and attribution on this page traces to one of the sources below, or is labeled as our own arithmetic. Entries marked "abstract only" were read as abstracts, so only what the abstract states is used from them. The Pearson pages were read on October 20, 2026.

• Stroop, J. R. Studies of interference in serial verbal reactions. Journal of Experimental Psychology, 18(6), 643 to 662, 1935. Text read at Classics in the History of Psychology on October 20, 2026.
• MacLeod, C. M. Half a century of research on the Stroop effect: An integrative review. Psychological Bulletin, 109(2), 163 to 203, 1991.
• Scarpina, F., and Tagini, S. The Stroop Color and Word Test. Frontiers in Psychology, 8, article 557, 2017.
• Checa, P., and Fernández-Berrocal, P. The role of intelligence quotient and emotional intelligence in cognitive control processes. Frontiers in Psychology, 6, article 1853, 2015.
• Arffa, S. The relationship of intelligence to executive function and non-executive function measures in a sample of average, above average, and gifted youth. Archives of Clinical Neuropsychology, 22(8), 969 to 978, 2007. Abstract only.
• Friedman, N. P., Miyake, A., Corley, R. P., Young, S. E., DeFries, J. C., and Hewitt, J. K. Not all executive functions are related to intelligence. Psychological Science, 17(2), 172 to 179, 2006. Abstract only.
• Friedman, N. P., and Miyake, A. Unity and diversity of executive functions: Individual differences as a window on cognitive structure. Cortex, 86, 186 to 204, 2017.
• Rey-Mermet, A., Gade, M., and Oberauer, K. Should we stop thinking about inhibition? Searching for individual and age differences in inhibition ability. Journal of Experimental Psychology: Learning, Memory, and Cognition, 44(4), 501 to 526, 2018. Abstract only.
• Kane, M. J., and Engle, R. W. Working-memory capacity and the control of attention: The contributions of goal neglect, response competition, and task set to Stroop interference. Journal of Experimental Psychology: General, 132(1), 47 to 70, 2003. Abstract only.
• Hedge, C., Powell, G., and Sumner, P. The reliability paradox: Why robust cognitive tasks do not produce reliable individual differences. Behavior Research Methods, 50(3), 1166 to 1186, 2018.
• Verhaeghen, P., and De Meersman, L. Aging and the Stroop effect: A meta-analysis. Psychology and Aging, 13(1), 120 to 126, 1998. Abstract only.
• Van der Elst, W., Van Boxtel, M. P. J., Van Breukelen, G. J. P., and Jolles, J. The Stroop Color-Word Test. Assessment, 13(1), 62 to 79, 2006. Abstract only.
• Karhadkar, M. A., Friedman, N. P., Reynolds, C. A., Corley, R. P., and Gustavson, D. E. Genetic and environmental associations between processing speed and executive functions across adolescence and established adulthood. Behavior Genetics, 56(4), 169 to 183, 2026. Abstract only.
• Pearson. Delis-Kaplan Executive Function System product page and WAIS-5 overview brochure. Both read on October 20, 2026.
• American Educational Research Association, American Psychological Association and National Council on Measurement in Education. Standards for Educational and Psychological Testing. AERA, 2014, Standards 1.1, 1.12, 1.14 and 2.3.
• American Psychological Association. Ethical Principles of Psychologists and Code of Conduct, Standards 9.02 and 9.06, 2017 text read in a public copy on October 20, 2026.

14 Frequently Asked Questions