IQ vs GMAT: Related, Not Equal

0 The Short Answer

The GMAT and a cognitive battery both make you reason under time pressure, so their scores move together across large groups, but no GMAT result can be translated into an IQ number. The reason is not squeamishness about labels. It is that the two instruments are scaled against different populations, cover different territory, and are built to answer different questions. One estimates how well a self selected pool of business school applicants is likely to handle a first year of graduate management coursework. The other describes where a person sits, across several separate cognitive domains, relative to adults in general. That second frame exists only when the comparison group is defined in advance, which is what an IQ test normed for adults supplies and what a scale built from business school applicants cannot.

There is also a practical trap specific to this exam. The GMAT you may remember is not the GMAT that exists now. The current version, which the Graduate Management Admission Council calls the Focus Edition, runs three sections: Quantitative Reasoning, Verbal Reasoning and Data Insights. Its Total Score runs from 205 to 805, not the 200 to 800 many people still quote from memory. Every conversion table circulating online that keys off a "700 GMAT" is describing a retired exam with different content and a different score distribution.

This page does three things the sibling comparisons on this site do not. It explains what item level computer adaptive testing actually does to a score, and why that makes percent correct intuition useless. It contrasts that adaptivity with the much cruder branching used inside professional cognitive batteries. And it takes Data Insights seriously as a measurement object, because it samples a competence that no classical intelligence battery has ever tried to assess in that form.

1 What the GMAT Is Now, After the Focus Redesign

Start with the physical facts, because most of the confusion downstream comes from people arguing about an exam that no longer exists. According to GMAC's official structure page, the current exam is 64 questions long and runs 2 hours and 15 minutes, with one optional 10 minute break that you may take after your first or your second section.

• Quantitative Reasoning. 21 questions in 45 minutes. All of them are Problem Solving items. No calculator is permitted here.
• Verbal Reasoning. 23 questions in 45 minutes, split between Reading Comprehension items that probe main idea, inference and logical structure, and Critical Reasoning items that ask you to strengthen an argument, weaken it, or name its flaw.
• Data Insights. 20 questions in 45 minutes, with an on screen calculator available. Five formats appear: Data Sufficiency, Multi Source Reasoning, Table Analysis, Graphics Interpretation and Two Part Analysis.

Two structural choices belong to the test taker. You pick the order in which you attempt the three sections. And within each section, once you reach the end and still have time on the clock, you may look back at as many questions as you want and change up to three of your answers. That editing allowance is capped, deliberately, and the cap is doing real psychometric work, which section 3 comes back to.

Notice what left the exam. The Analytical Writing Assessment no longer contributes to the score you send to schools. Sentence Correction, once a fixture of the verbal section, is gone. Geometry, once a reliable slice of the quantitative content, is gone. And Data Sufficiency, historically a quantitative item type, now sits inside Data Insights alongside charts and multi tab source documents. That last move is not cosmetic. It reclassifies a reasoning format as a data handling format, which tells you something about how GMAC now thinks about what business schools need to know.

The arithmetic of the clock matters for the trainability discussion later. Quantitative Reasoning allows roughly 2 minutes and 9 seconds per question. Verbal Reasoning allows about 1 minute and 57 seconds. Data Insights allows exactly 2 minutes and 15 seconds. Those budgets are tight enough that pacing is a skill in its own right, separable from whether you could solve the same items with unlimited time.

2 Reading the 205 to 805 Scale Without Fooling Yourself

The Total Score runs from 205 to 805, and every reportable value ends in a 5, so the scale advances in steps of ten. Each of the three sections is reported separately on a 60 to 90 scale in single point steps, and GMAC states that all three contribute equally to the Total Score. Results stay valid for five years. Every admissions scale is arbitrary in this same way: the LSAT's 120 to 180 band exists to keep scores equivalent across test forms and bears no relationship whatever to a mean of 100 and a standard deviation of 15.

Equal weighting is a bigger deal than it sounds. On the retired exam, the headline number came from two sections, and quantitative performance dominated the conversation among applicants. Now a third of your Total Score comes from a section devoted to reading tables and charts. Someone with a strong mathematics background and no habit of interrogating a data display can lose a third of the available signal, which was much harder to do before.

The score distribution is the part people consistently get wrong. Because the old scale topped out at 800 and a 700 had folkloric status, applicants transplant that intuition onto a scale where the numbers mean something else. GMAC's own score documentation says outright that comparing Total Scores between the current edition and the previous one is not an appropriate or meaningful comparison, and directs readers to compare percentile rankings instead. Here is what those percentiles look like, taken from the concordance table GMAC published in August 2026 using test taker data from July 2021 through June 2026.

Hold on to the 705 figure. It reappears in section 10, where it does most of the work of showing why percentile matching is the single most common way people talk themselves into a fake conversion.

3 What "Computer Adaptive" Actually Does to a Score

GMAC describes the mechanism plainly: the exam adapts the questions it shows you based on a real time assessment of how you handled earlier ones. Answer correctly and the next item tends to be harder. Answer incorrectly and the next one tends to be easier. That description is accurate but it undersells what is happening underneath, and the underneath is where the comparison with cognitive testing gets interesting.

Behind the exam sits a large bank of items, each of which has been pretested on real candidates and assigned calibrated statistical properties, principally a difficulty value on a common scale. The algorithm maintains a running estimate of your proficiency and, at each step, selects an item whose difficulty sits close to that current estimate. Every response tightens or shifts the estimate, and the next selection follows the update.

The reason this improves precision is easy to state and easy to forget. An item tells you the most about a person when its difficulty is near their ability, because that is the point where the outcome is genuinely uncertain. An item far below someone's level is answered correctly by almost everyone at that level and carries almost no information. An item far above it is missed by almost everyone and carries almost as little. A fixed form has to serve the whole ability range with one set of questions, so most test takers spend most of the test on items that are barely informative for them personally. Adaptive selection spends nearly every item in the informative zone.

The efficiency gain is measurable. In simulation work on adaptive reading assessment published in the Journal of Computer Assisted Learning in 2024, Ebenbeck and colleagues found that an estimation based adaptive procedure cut test length by around 40 percent while retaining an average accuracy of r = .96 against the full length benchmark. In a 2024 experimental study in Personality and Individual Differences, Akhtar and colleagues randomly assigned 286 participants to an adaptive or a fixed item version of a multidimensional fluid reasoning test and found the adaptive version delivered better measurement precision. That study also caught two effects worth knowing: participants in the fixed item condition believed they had answered more items correctly, and rapid guessing rose as the test went on, particularly in the fixed condition when the hardest items arrived.

Three consequences follow for anyone trying to reason about a GMAT score.

• Number correct is not a score. Two candidates who each get 45 of 64 right can receive very different Total Scores, because the difficulty of what they got right differs. Any chart claiming to map raw correct answers to a Total Score is describing a test that does not work this way.
• Precision is roughly even across the range. Fixed forms measure the middle of the distribution well and the extremes badly, because they run out of appropriately hard or easy items. Adaptive selection keeps feeding informative items at every level, so the reported score is about as trustworthy at 745 as at 545.
• The three answer edit is a bounded concession. By the time you reach the review screen, the items you saw were already chosen off your original responses. Editing changes what gets scored; it cannot change the path the algorithm walked. Capping the edits at three keeps that mismatch small enough to stay tolerable.

4 Where GMAT Adaptivity and Adaptive IQ Testing Part Ways

Cognitive batteries are adaptive too, and people who learn this often assume the two designs are cousins. They are not, and the differences illuminate what each instrument is actually for. Adaptive delivery is now common in browser based assessment as well, and it is one of several design details worth checking before trusting a normed IQ test taken from home, alongside its reference group and its published reliability.

The individually administered batteries used by psychologists adapt by block, not by item. A subtest has age based start points, so a 45 year old does not begin at item one. It has reversal rules, so a stumble at the start sends the examiner backwards until a floor is established. And it has discontinue rules, so the subtest stops after a run of consecutive failures rather than marching the person through items far above their reach. That is real adaptation, and it exists for the same reason the GMAT's does: to stop wasting the test taker's attention on uninformative items. But it is coarse, human operated branching, and it does not recompute a proficiency estimate after every response.

Four differences run deeper than the mechanics.

• One dimension per section versus a deliberate profile. Each GMAT section adapts along a single proficiency estimate, because each section is designed to yield one number. A cognitive battery is built to resist collapsing into one number: it separates verbal comprehension, fluid reasoning, visual spatial processing, working memory, processing speed and quantitative reasoning precisely because those domains come apart within a person. Adaptivity there has to run separately inside each domain or it destroys the thing being measured.
• Fixed content is a feature on the norming side. The GMAT's bank is huge and continuously refreshed, partly for security: if everyone saw the same 64 items, the items would leak. A norm referenced battery usually keeps a fixed item set so that every person's raw performance maps onto the same standardization table. Refreshing the items would break the norms, which is exactly the property that lets a score be interpreted as a position in a population.
• Different precision priorities. The GMAT wants sharp discrimination in the crowded upper middle of an applicant pool, because that is where admissions decisions get made. A cognitive battery has to behave sensibly from severe impairment to high giftedness across a wide age span, which is why floor and ceiling coverage dominates its design. Those goals pull the engineering in opposite directions.
• Different scaling targets. An adaptive estimate is meaningless until it is scaled against somebody. The GMAT scales against people who chose to sit the GMAT. An IQ scale sets its mean at 100 with a standard deviation of 15 against an age representative sample, which is a completely different reference object. The norming procedure is where the meaning lives, and adaptivity does not change that.

The honest summary: the GMAT's adaptive engine is more sophisticated, item for item, than what runs inside a standard clinical battery. It is also aimed at a narrower target. Sophistication in the algorithm does not widen the construct being estimated.

5 Data Insights: The Section With No Cognitive Battery Equivalent

Data Insights is the most genuinely novel thing on the current exam, and it is the reason this comparison is more interesting than the equivalent comparison for a law or graduate school entrance test. GMAC describes it as measuring the ability to analyze and interpret data for business decisions, including digital and data literacy. In practice it asks you to do things that appear nowhere in a classical intelligence battery.

Consider what Multi Source Reasoning demands. Information sits across several tabs: a memo, a table, maybe a chart. No single tab answers the question. You have to decide which tab is even relevant, hold a partial result from one while you read another, notice that the memo's date range does not match the table's, and resolve the conflict. Graphics Interpretation asks you to read quantity out of position, length or area, then complete a sentence with a dropdown, which means you must also recognize which verbal claim your visual reading supports. Table Analysis gives you a sortable table and asks a series of yes or no judgments, each of which needs a different sort to answer efficiently.

Map that onto CHC constructs and the mixture is unmistakable. Extracting magnitude from a bar or a scatter recruits visual spatial processing. Holding a subtotal from tab one while parsing tab two recruits working memory. The arithmetic recruits quantitative reasoning. Deciding which of three plausible readings the footnote licenses recruits verbal comprehension. And knowing that a truncated y axis exaggerates a difference, or that a percentage of a percentage is not a percentage, is crystallized knowledge of graphical convention, learned rather than derived. No subtest in a standard battery presents that compound. Batteries deliberately isolate domains so that a profile means something; Data Insights deliberately fuses them because business decisions arrive fused.

This is not a criticism of either design. It is a statement about non overlap. And there is evidence that the underlying competence is partly taught rather than purely derived from reasoning capacity. In a 2025 study in Cognitive Research: Principles and Implications, Brockbank and colleagues gave two widely used graph comprehension assessments to 1,113 adults across a university sample and a demographically representative sample. Overall scores on the two instruments correlated, and both tracked how much prior mathematics coursework participants had. But the error patterns suggested the two assessments probe somewhat distinct components of data visualization literacy, and the components did not line up with the categories their designers had assumed. If the field's own dedicated measures of this skill do not fully agree with each other, treating a Data Insights score as a clean readout of general reasoning is not defensible.

One more detail carries a design message. The calculator is available in Data Insights and forbidden in Quantitative Reasoning. GMAC is saying, as clearly as a rulebook can, that the quantitative section wants to see whether you can compute and the data section does not care. Data Insights is testing what you decide to compute and whether you believe the answer, which is a different question entirely.

6 Quantitative Reasoning Is a Narrow Cut of Quantitative Ability

The current quantitative section is 21 Problem Solving items in 45 minutes, testing what GMAC calls algebraic and arithmetic foundational knowledge and its application. That description is narrower than it used to be, and the narrowing is informative. Geometry left. Data Sufficiency moved out. What remains is arithmetic, algebra, word problems, rates, ratios, statistics of the mean and median variety, and the ability to see through a wordy setup to the equation hiding inside it.

Compare that with how quantitative reasoning is assessed inside a cognitive battery. There the aim is to separate what you have learned from what you can derive. Some items deliberately use minimal content so that a person who never finished secondary mathematics can still demonstrate reasoning with quantity. Others test acquired mathematical knowledge on purpose, and the two are kept apart so the profile can show a gap between them. The GMAT has no interest in that separation, because a business school does not care whether your quantitative competence is native or taught. It cares whether it is present in September.

The practical consequence is that content recency drives a large share of the observed variance in this section. An applicant three years out of an engineering degree and an applicant twelve years out of a humanities degree can have identical fluid reasoning and produce quantitative scores far apart, because one of them has been rehearsing algebraic manipulation continuously and the other has not touched it since adolescence. On a battery, that gap would show up as a discrepancy between the acquired knowledge index and the fluid reasoning index, and the report would say so. On the GMAT it shows up as a single lower number with no explanation attached.

This is the clearest single reason why an individual's quantitative section score cannot stand in for anything IQ shaped. The score is a compound of reasoning capacity, curriculum history, recency of practice and pacing under a two minute budget, and the exam publishes no decomposition of that compound because it does not need one.

7 Verbal Reasoning Is Reading Comprehension Under a Stopwatch

Twenty three items in 45 minutes, drawn from two types. Reading Comprehension asks for main idea, inference and logical structure from dense passages. Critical Reasoning presents a short argument and asks what would strengthen it, what would weaken it, or where the reasoning breaks. Spotting a flaw when a question announces that one is present is a different achievement from noticing one unprompted, and the research on how thinking well and scoring well diverge found that susceptibility to the conjunction fallacy, anchoring and base rate neglect had little to no association with measured ability across seven studies by Stanovich and West.

The removal of Sentence Correction matters for anyone comparing this with a verbal comprehension index. Sentence Correction rewarded knowledge of formal written English conventions, which is a vocabulary and usage competence. Without it, the section is almost entirely about extracting and evaluating propositional content. That actually moves the section closer to what a fluid reasoning task does, and further from what a vocabulary subtest does, which is the reverse of what most people assume when they hear "verbal".

What it does not remove is language background. Reading three dense passages and eleven argument stems in under 45 minutes is a reading rate problem before it is a reasoning problem. A candidate who reasons about arguments beautifully in their first language and reads academic English at 180 words per minute will lose items to the clock that they would answer correctly with two more minutes each. A verbal comprehension index built on orally administered items, with no reading load and no comparable time pressure, would not register that loss at all. The two measures diverge for a reason that has nothing to do with the construct either is trying to capture.

Critical Reasoning has a second property worth naming: its question stems recur. There is a finite catalogue of argument flaws and a finite set of ways to ask about them. A candidate who has worked through several hundred of these items is not merely faster, they are pattern matching against a memorized taxonomy. That is legitimate preparation and it is exactly what the exam intends to reward, since business school involves reading arguments made by interested parties. It is also, straightforwardly, learned content, which is the property that makes the section unsuitable as a proxy for anything.

8 Side by Side: Two Instruments Built for Different Jobs

Laying the design decisions next to each other makes the incompatibility concrete. Nothing in the right column is better than the left; they are answers to different questions.

Read down the last two rows in particular. An instrument validated against first year grades has earned the right to make claims about first year grades. Extending that to claims about a person's cognitive architecture is not a small stretch of the evidence, it is a change of subject. The right hand column is also not one product but a shelf, and the major batteries differ by publisher, edition, age range and administration mode, so which one a report came from already decides what its numbers can be set beside.

9 The Norm Group Is the Whole Argument

Everything that makes GMAT to IQ conversion fail traces back to one fact: percentiles are meaningless without their denominator, and these two instruments use denominators that barely overlap. Denominators also move underneath you: the ACT's national ranks are rebuilt as each cohort of recent graduates changes, so an unchanged scaled score can come back with a slightly different percentile in a later reporting period.

Who sits the GMAT? People who have decided to pursue graduate management education, who have almost always completed an undergraduate degree, who are willing to pay a substantial registration fee, and who typically prepare for weeks or months beforehand. That is a triple filter of education, motivation and money. Who is in an IQ standardization sample? People recruited to mirror the general adult population on age, sex, education and region, most of whom did not choose to be measured and had no reason to prepare.

Selecting on ability compresses the spread of ability, and compressed spread mechanically shrinks correlations. That is the restriction of range problem, and the research literature has spent two decades arguing about how much to correct for it. Sackett and colleagues, writing in the Journal of Applied Psychology in 2021, reexamined the standard corrections used across personnel selection and concluded that the usual artifact distribution methods systematically overcorrect, cutting mean validity estimates by .10 to .20 once fixed. Steel and colleagues followed in the same journal in 2024, showing that using national norms rather than the actual applicant pool variance inflates corrections badly, and that in modern applicant pools range restriction corrections turn out to be minimal in around three quarters of cases, because education itself has already done the filtering.

The lesson generalizes past personnel selection. Any corrected coefficient you meet in this area should be read as an upper bound produced by a contested procedure, not as a measurement. And any percentile you meet should be read together with its population, or it is a decoration. The limiting case is an instrument that publishes no reference population at all, which is why an ICAR raw score cannot honestly be reported as a percentile despite ICAR being a serious public domain research instrument.

Here is the concrete version. A Total Score of 705 sits at the 97.5th percentile among GMAT test takers in GMAC's 2021 to 2026 reference window. An IQ of 130 sits at roughly the 98th percentile of the general adult population, which is about 1 person in 44. Those two percentiles look almost identical, and the temptation to write "705 equals about 130" is enormous. It is also wrong, because the 97.5th percentile of a filtered pool of degree holding, motivated, prepared candidates is a much rarer position in the general population than the 98th percentile of everyone. Matching percentiles across incompatible reference groups is the single most common conversion error, and it is the one that feels most rigorous while you are making it. Our percentile reference chart shows how the general population version of this ladder behaves.

10 Why GMAT to IQ Conversions Have Nothing Underneath Them

The strongest argument against these conversions is not philosophical. It is that GMAC cannot cleanly convert the GMAT into the GMAT. The proof carries over unchanged to the GRE, which shares the applicant pool norm group and the coachable content that break any conversion.

When the Focus Edition replaced the previous exam, schools needed a way to compare applicants holding scores on two different scales, so GMAC published a concordance table. That document does exactly what a defensible conversion looks like: it names both instruments, states the data window used to build it, and is republished annually. Its August 2026 edition draws on test taker data from July 2021 through June 2026. And here is what it shows. A Total Score of 700 on the previous edition maps to 645 or 655 on the current one. A 750 maps to 695, 705 or 715. The mapping is a range, not a point, and GMAC explains why: percentile computation involves differences in score bin sizes and observed frequencies, so linked scores legitimately span a band.

Sit with that. Two editions of the same exam, published by the same organization, taken by overlapping populations, measuring content that is largely continuous, and the honest mapping between them is still fuzzy by two score bins. GMAC states elsewhere that comparing Total Scores across editions is not an appropriate or accurate comparison at all, and recommends comparing percentile ranks instead. If that is the level of precision available between two versions of one test, a formula purporting to convert a GMAT Total Score into a Full Scale IQ, across different constructs, different populations, different scales and different eras, is not a weaker version of the same thing. It is a different kind of object entirely, and the difference is that nothing was measured.

Three further problems compound it.

• No joint dataset exists. A defensible conversion requires the same people to take both instruments, with both scores reported, in a sample large enough to estimate the relationship and a second sample to validate it. The published GMAT validity literature reports the exam against graduate grade point average, not against Wechsler style indices. There is no equivalent of the Koenig and Detterman work that pinned the ACT to g using the National Longitudinal Survey of Youth, and until such a study exists on the Focus Edition specifically, every published GMAT to IQ figure is an extrapolation from an extrapolation.
• Regression to the mean bites hardest at the top. Two imperfectly correlated measures always pull extreme scores back toward average. Someone at the 99th percentile on the GMAT should be expected to land closer to the middle on a second, different instrument, and the more extreme the first score the larger the expected pull. Conversion tables are typically built and marketed at exactly the high end where this effect is strongest.
• The composite hides the profile. Two candidates can both score 665 with completely different section splits, one carried by Quantitative Reasoning and one by Data Insights. If a single number cannot distinguish those two candidates from each other, it cannot possibly resolve a six domain cognitive profile. This is the same reason a Full Scale composite is the least informative line on a well written score report.

11 What Preparation Moves, and What It Does Not

The GMAT is a coachable exam and nobody at GMAC pretends otherwise, since the organization sells official preparation material. Being specific about which parts move is more useful than a general warning. Coachability is also why a score can be stable and still not measure what it claims, since validity, reliability and a defined norm group are what separates an accurate IQ score from a guess.

Highly responsive to study. Arithmetic and algebra content that has gone stale, which is pure recall and comes back quickly. The catalogue of Critical Reasoning question stems, which is a finite taxonomy. Fluency with the five Data Insights formats, especially the mechanics of sorting a table or reading a dual axis chart, where the first exposure costs a minute and the twentieth costs ten seconds. Pacing against a two minute budget, which is a rehearsable behavior. Section order strategy, which costs nothing to optimize. And familiarity with the adaptive experience itself, since knowing that difficult items are a sign the algorithm has revised its estimate upward prevents the mid section panic that wrecks pacing.

Barely responsive. The reasoning capacity that determines how quickly you see the structure of an unfamiliar problem. Working memory span. Reading rate in a second language, which improves over years rather than weeks.

It is worth noting that the popular picture of coaching may be inflated. Dahlke and colleagues, publishing in Applied Measurement in Education in 2023, tracked 120,384 students across low stakes and high stakes administrations of the PSAT and SAT from ninth through twelfth grade. If expensive coaching produced routine large gains, high socioeconomic status students should have shown a burst of improvement concentrated on the high stakes attempt. Instead, score improvement looked similar in both settings, with 3.4 percent of high status and 1.1 percent of low status students showing larger than expected gains. The gap is real but small, and inconsistent with the idea that coaching reliably transforms scores.

The asymmetry that matters for this page: a Total Score can plausibly move 100 points with disciplined preparation, and that movement is a valid reflection of improved readiness for the coursework the exam predicts. There is no comparable published evidence that studying business problems moves a full scale cognitive index by an equivalent amount. Practice effects on cognitive batteries at retest are real, modest, and largely tied to the specific instrument rather than transferring across measures. So if a person's GMAT rises by 100 points and their cognitive profile does not shift, nothing paradoxical has happened. The two numbers were tracking different things all along, which is what section 12 is about.

12 What Each Score Actually Predicts

Ask what an instrument forecasts and the comparison stops being abstract.

The GMAT's evidence base is about graduate management performance, and it is substantial. Kuncel, Credé and Thomas published a meta-analysis in Academy of Management Learning and Education in 2007 aggregating the predictive validity of GMAT scores and undergraduate grade point average for graduate student academic performance. Oh and colleagues revisited that same database in the same journal in 2008 using an improved range restriction correction and concluded that the earlier work had understated the exam's validity by around 7 percent, which is a useful reminder that correction methodology moves these numbers as much as the data does. Sireci and colleagues, writing in Educational and Psychological Measurement in 2006, analyzed admissions and first year grade data from 11 graduate management schools and found that GMAT verbal and quantitative scores accounted for roughly 16 percent of the variance in graduate grade point average beyond what undergraduate grades predicted, with the two sources together reaching about 25 percent. That study also found no statistical differences in prediction across groups defined by sex or race and ethnicity, and found the analytical writing score contributed only about 1 percent, which is presumably part of why it no longer counts toward the Total Score.

Talento-Miller and colleagues, in the same journal in 2008, summarized 273 validity studies conducted between 1997 and 2004 and reported an interquartile range of .45 to .63 for GMAT scores combined with undergraduate grades predicting early graduate performance. Read those figures alongside the overcorrection findings from section 9 and treat them as a well supported band rather than a precise constant.

A cognitive battery's evidence base points elsewhere. Broad measured ability relates to training success and performance across a wide span of occupations, to educational attainment, and to the acquisition of complex skills over long horizons. It says nothing specific about how you will handle a case method classroom, a quantitative finance core, or the particular reading load of an MBA program. Standardized admission tests are reasonable proxies for general ability in aggregate research, as Wai and colleagues argued in Journal of Intelligence in 2018, and Coyle and Pillow showed in Intelligence in 2008 that the SAT and ACT remain predictive of college grades even after their g variance is removed, meaning the non g content in these tests is doing real work rather than being noise.

Put plainly: the GMAT predicts a specific thing well and was engineered to. A cognitive profile describes a broad thing usefully and was engineered to. Asking either to do the other's job produces a worse answer than either gives on its own. If what you want is a domain by domain picture rather than an admissions number, that is a different measurement and it should be taken as one. ACIS reports a profile across six CHC domains with intervals around each, which is a description of relative strengths, not a credential and not a diagnosis. What a cognitive score does and does not capture is worth understanding before you take any of them seriously.

13 Frequently Asked Questions

14 Sources Behind This Page

Comparing tests means comparing what each one measures. The sources below cover the official test documentation and the research on how admissions scores track cognitive ability.

• Graduate Management Admission Council. The official GMAT documentation: sections, scoring and score reporting.
• Kuncel, N.R., Hezlett, S.A. & Ones, D.S. (2004). Academic performance, career potential, creativity, and job performance: can one construct predict them all? Journal of Personality and Social Psychology, 86(1).
• Frey, M.C. & Detterman, D.K. (2004). Scholastic Assessment or g? Psychological Science, 15(6). The study linking SAT scores to measured cognitive ability, and the reason SAT conversions are estimates rather than measurements.
• Koenig, K.A., Frey, M.C. & Detterman, D.K. (2008). ACT and general cognitive ability. Intelligence, 36(2). The companion study on how ACT scores track g.
• Study of Mathematically Precocious Youth, Vanderbilt University. A 50 year longitudinal study of five cohorts totalling over 5,000 individuals, identified between the top 3% and the top 0.01% of ability and followed for decades.
• Voncken, L., Albers, C.J. & Timmerman, M.E. (2019). Improving confidence intervals for normed test scores. Behavior Research Methods. Open access. Documents the mean 100, SD 15 metric and the uncertainty that norming from samples adds to any score.
• Crawford, J.R., Garthwaite, P.H. & Slick, D.J. (2009). On percentile norms in neuropsychology. The Clinical Neuropsychologist, 23(7), 1173-1195. Three definitions of a percentile coexist in practice and can return different results for the same score.
• Pearson Clinical Assessment Scientific Council (2023). Standardized Clinical Assessment for Practitioners: A Primer. How standard scores, percentile ranks and the standard error of measurement are meant to be read together.
• Pearson (2008). WAIS-IV Score Report sample. What a real report contains: every composite paired with a percentile rank, a 95% confidence interval and a qualitative description, never a bare number.
• Trahan, L.H., Stuebing, K.K., Hiscock, M.K. & Fletcher, J.M. (2014). The Flynn effect: A meta-analysis. Psychological Bulletin. Open access. Scores rose about 2.31 points per decade across 285 studies, which is why norms age and older scores overstate standing.