IQ vs the bar exam: what each part measures, what predicts passing and why no score converts

0 The short answer

The bar exam is a licensing test of acquired legal knowledge and lawyering skills, so no bar score converts to an IQ, and no study reviewed here measured IQ alongside bar results. Its scores sit on scales that are equated across sittings and not tied to a norm group of adults, a 400 point scale for the legacy exam and a 500 to 750 scale for NextGen, and each jurisdiction sets its own pass mark. In the studies reviewed here, law school grades and the LSAT both predict passing, grades more strongly once a student is enrolled, and much of the outcome stays unexplained. General ability probably contributes through who reaches law school and how fast the law is learned, but that is an interpretation because none of the studies measured it, and preparation and persistence move bar results a great deal.

1 What Is the Bar Exam Designed to Measure?

The bar exam is designed to decide whether a candidate has the minimum knowledge and skills to be licensed, which makes it a credentialing test and not an intelligence test. The National Conference of Bar Examiners, the nonprofit that develops the exam content, describes the Uniform Bar Examination as designed to test the knowledge and skills that every lawyer should be able to demonstrate before being licensed. Its Testing Task Force, in the final report of a three year study published in April 2021, put the purpose more bluntly: the exam exists to protect the public by helping to ensure that newly licensed lawyers possess the minimum knowledge and skills to perform activities typically required of an entry-level lawyer. Nothing in that purpose refers to a general capacity, a population of adults or a standard deviation.

That purpose shapes much about the score. The Standards for Educational and Psychological Testing, published jointly by AERA, APA and NCME in 2014, give licensing and certification their own treatment in chapter 11. Tests used in credentialing, the chapter says, are designed to determine whether the essential knowledge and skills have been mastered by the candidate. Their validity depends mainly on content evidence, meaning judgments that the test represents the content domain of the occupation, and criterion evidence is of limited use because a credentialing exam is not meant to predict individual performance in a specific job. The chapter adds that scores need to be precise in the vicinity of the cut score and may not need to be as precise for people who clearly pass or clearly fail.

An IQ test is built for a different job. It estimates a person's standing on general ability, and on broad abilities within it, relative to a reference group of adults, which is why its scale has a mean of 100 and a standard deviation of 15. The page on what IQ measures explains that design, and the page on how IQ scores are normed explains how the reference group fixes the scale. None of the sources reviewed describes a reference population of adults, or a scale tied to one, for the bar exam. It has a defined body of legal content, a pass mark, and a requirement that the pass mark mean the same performance in July as in February.

The two kinds of test are less far apart than that contrast suggests, and the difference between them is one of purpose and breadth. Kuncel and Hezlett, in a 2010 review in Current Directions in Psychological Science, describe standardized measures of intelligence, ability and achievement as all measuring acquired knowledge and skill. A bar exam sits at the narrow, content-specific end of that range, where a defined body of law is the target. A cognitive battery sits at the broad end, where the target is a spread of abilities such as fluid reasoning (Gf), crystallized knowledge (Gc), working memory (Gwm) and processing speed (Gs) and not one body of law. The page on what intelligence is defines the construct, the page on intelligence versus knowledge draws the same line in general terms, and the page on aptitude versus intelligence covers the middle ground where admission tests fall.

2 How Is the Legacy Bar Exam Built: MBE, MEE and MPT?

The legacy Uniform Bar Examination combines a 200 question multiple-choice test, six essays and two performance tasks, weighted 50, 30 and 20 percent into one score on a 400 point scale. NCBE describes the exam as uniformly administered, graded and scored, with a portable score that can be transferred between participating jurisdictions. The table below summarizes the three components from NCBE's own pages, read on October 6, 2026.

The seven MBE subjects, according to NCBE's guide to preparing for the MBE, are Civil Procedure, Constitutional Law, Contracts, Criminal Law and Procedure, Evidence, Real Property and Torts. Each question has four answer options, scores are based on the number answered correctly, and nothing is subtracted for wrong answers. The unscored pretest items cannot be told apart from scored ones, so an examinee has to treat all 200 as real. Two three-hour sessions of 100 questions each give an average of 1.8 minutes per question, our arithmetic on the published figures. The MEE and MPT together add six hours, so the legacy exam is 12 hours of testing spread across two days, again our arithmetic on the published component times.

NCBE has produced the MBE since 1972, according to its MBE statistics, and the Uniform Bar Examination, which packages the three components, was first administered in February 2011, according to its UBE statistics. The division of labor matters for the score. NCBE scores the MBE. Jurisdictions grade the MEE and the MPT, and NCBE then scales those written scores to the MBE and calculates the total. Jurisdictions also decide who may sit, how many times, what the passing score is, whether a local law component is added, and how long a transferred score stays valid. NCBE reports that its exam content is used by 54 United States jurisdictions, according to its press release of September 25, 2026.

3 What Does Each Part of the Bar Exam Ask of the Examinee?

Each part of the bar exam asks for a different mix of legal knowledge, reading, reasoning and writing, and only some of that mix is what a cognitive battery calls general ability. The mapping below is our reading of NCBE's published descriptions in the vocabulary of the CHC model, which names the broad abilities that cognitive batteries sample: Gc is crystallized knowledge, Gf fluid reasoning, Grw reading and writing, Gwm working memory, Gs processing speed, Gv visual processing and Gq quantitative knowledge. NCBE has not published a factor analysis of the bar exam against a cognitive battery, and none of the sources reviewed for this page did either, so the table is a hypothesis about task demands and not a measured loading.

The MBE is mostly acquired knowledge in a narrow domain. Gc, the crystallized side of the CHC model, is knowledge and the skill of using it, and the page on fluid versus crystallized intelligence explains why it grows with exposure and study. A candidate who has never studied the law of evidence cannot reason to the right answer on an evidence question, however strong their fluid reasoning, because the rule being applied is not on the page. The reasoning component is real, since answer options are built so that near-miss rules are tempting, but it operates on rules held in memory. A Texas Tech study discussed below shows the knowledge side directly. Grades in seven law courses matching the tested subjects predicted 35.6 percent of the variance in the MBE total for the 190 examinees in 2014 who had only a composite score, and only Contracts and Evidence added independent weight, according to Austin, Christopher and Dickerson in the Hofstra Law Review.

The MEE adds writing. NCBE says the primary distinction between the MEE and the MBE is that the MEE requires the examinee to demonstrate an ability to communicate effectively in writing, and jurisdictions, not NCBE, grade the essays. The MPT is the part closest to a reasoning test, because the law and the facts are both supplied. NCBE's guide to the MPT says that the file includes irrelevant facts, that facts can be ambiguous, incomplete or even conflicting, and that the library may contain authorities that are not relevant. The examinee has to sort, reason from the authorities to the facts and write the result inside 90 minutes. In CHC terms that leans on fluid reasoning, reading and writing, and working memory, with a clock that adds speed. The Testing Task Force's listening sessions found the MPT widely viewed as the component most representative of the skills that newly licensed lawyers need on entering practice.

The job analysis behind the newer exam gives a second view of what is being sampled. The Testing Task Force's practice analysis was completed by 14,846 lawyers, 3,153 of them newly licensed, and the five skills, abilities and other characteristics with the highest mean criticality, out of the 36 rated, were written and reading comprehension, critical and analytical thinking, written expression, identifying issues, and integrity and honesty.

The exam also leaves out most of what a Full Scale IQ test samples across the six cognitive domains: no part is designed to measure spatial ability (Gv), the span of working memory (Gwm) or the speed of simple decisions (Gs), and quantitative reasoning (Gq), the subject of the page on the quantitative reasoning test, is not a tested content area, which suits a licensing exam, and what it does cover leans on knowledge acquired through study.

4 What Changes With the NextGen Bar Exam?

The NextGen bar exam, first administered in July 2026, replaces three separately scored components with one integrated nine hour exam that reports a single score from 500 to 750. NCBE's announcement of July 31, 2026 reported that the inaugural administration was held that week in 10 jurisdictions, with 2,640 examinees completing it out of about 3,318 who registered. The legacy exam continues through February 2028, and NCBE expects 52 jurisdictions to begin administering the NextGen exam between that announcement and July 2028.

The design is described on NCBE's NextGen pages. The exam runs over one and a half days, two three-hour sessions on the first day and one on the second, on the examinee's own laptop at a proctored site. The sample questions page says that about 40 percent of exam time goes to standalone multiple-choice questions, just over a quarter to integrated question sets built on a common fact scenario with legal resources and supplemental documents such as a police report, and about a third to performance tasks, which NCBE compares to the legacy MPT.

Two features bear on the question of ability. First, other areas of law, including family law and trusts and estates through at least February 2028, appear only in questions that supply the legal resources, and NCBE says examinees are not expected to build a base of knowledge in them. Second, the Testing Task Force reported that both of its committees agreed that doctrine should be limited to the core principles that new lawyers need to know without looking them up, not the exceptions to the exceptions. Our reading is that this shifts weight from memorized rules toward reading, reasoning and writing with supplied materials. NCBE reports that early survey responses frequently highlighted practical skills, realistic tasks and applying legal knowledge rather than relying primarily on memorization. No published analysis that this review found relates NextGen scores to a cognitive battery, so the shift in weight is an expectation and not a result.

NCBE released the first national results on September 25, 2026. In its press release, 2,630 examinees earned an official score, the national mean was 640.6, the median was 640, and the overall pass rate was 72.7 percent, within 4.3 percentage points of every July national pass rate from 2021 through 2025. First-time examinees passed at 85.2 percent and repeat examinees at 41.0 percent, although NCBE cautions that first-time now means a first bar exam in any jurisdiction, a stricter definition than before. NCBE also stresses that 10 jurisdictions and a smaller group of examinees cannot answer every question about performance across populations.

The NextGen scale is a new ruler. NCBE's score guidance says the legacy and NextGen exams use different score structures and scales, so numerical scores should not be compared directly, and that most jurisdictions have set NextGen passing scores between 610 and 620. NextGen scores are equated and not curved to the group that sits together. There is also no separate multiple-choice and written score, so the legacy habit of reading an MBE number as a measure of knowledge alone has no NextGen counterpart.

5 How Does a Bar Score Get Its Number?

A bar score is built in three steps, equating the multiple-choice test, scaling the written scores to it and applying a state pass mark, and none of the three ties the number to the general population. The steps explain why a bar score cannot be read the way an IQ is read, so they are worth taking one at a time.

The first step is equating. NCBE's page on MBE scores explains that a raw score is the number of questions answered correctly, that raw scores from different administrations are not comparable because the questions differ in difficulty, and that equating adjusts for this so that a scaled score represents the same level of performance on any test date. The scaled score ranges from about 40 to 200. The method rests on repeated items. A Bar Examiner explanation by Susan Case, reprinted in The Testing Column of Fall 2020, describes a mini-test of items from earlier exams embedded in the larger exam, which lets NCBE compare this administration's candidates with a previous group on exactly the same items. Equating therefore removes differences in question difficulty. It does not remove differences between groups of examinees. NCBE's 2025 MBE statistics show a national mean of 142.4 in July, with a standard deviation of 17.9, and 130.8 in February, with a standard deviation of 15.6. The scale is stable across sittings while the groups that sit differ.

The second step is scaling the written scores. Essay and performance questions are not reused, so they cannot be equated. NCBE instead rescales a jurisdiction's written scores so that they have the same mean and standard deviation as that jurisdiction's scaled MBE scores, which preserves each examinee's rank order. The Fall 2020 Testing Column gives the reason: average essay scores in February tend to match average essay scores in July even though February candidates are consistently less proficient as a group, so raw essay scores float with the pool. NCBE's 2025 UBE statistics show the result on the 400 point scale, a national mean of 285.5 in July and 260.0 in February, with standard deviations of 34.2 and 30.2, and an overall mean of 278.6 with a standard deviation of 35.0.

The third step is the pass mark, and it is a policy choice. NCBE's table of minimum scores, read on October 6, 2026, lists 260 in eight jurisdictions, 264 in Indiana, 266 in 13, 268 in Michigan and 270 in 18. Jurisdictions outside the Uniform Bar Examination use other scales. NCBE's July 2026 results table lists passing scores of 1,390 for California, 136 for Florida, 143 for Delaware and 75 for Nevada in its non-UBE rows. A UBE score of 266 passes in New York, where the minimum is 266, and falls short in Texas, where it is 270. The gap between the lowest and highest legacy UBE pass marks, 10 points, equals about 0.29 of the overall 2025 standard deviation of 35.0, our arithmetic. The Standards say that the level required to pass should depend on the knowledge and skills needed for credential-worthy performance and should not be adjusted to control the number of people who pass.

Put the three steps together and the scale has a clear meaning. A bar score says how a candidate's performance compares with a standard that is held constant across sittings, on a scale whose center sits where bar takers have scored, with a pass mark chosen by a state. NCBE describes the scale in terms of equating across administrations, and none of the sources reviewed describes norming it against adults in general. A bar score therefore does not say where the candidate stands among adults. That is the opposite design from an IQ, whose scale exists to place a person among adults, and the page on the standard deviation of 15 shows what that placement means.

6 What Predicts Passing the Bar?

Across a national cohort and three later studies, law school grades and the LSAT predict passing the bar, grades more strongly once a student is enrolled, and much of the outcome stays unexplained. None of the studies included an intelligence test, so they show how admission credentials and law school performance relate to the exam, not how IQ relates to it. The page on IQ and the LSAT compares that admission test with cognitive batteries. This page takes the LSAT only as one predictor of the bar.

The national study is the largest and the oldest. It followed students through up to five bar administrations, and when law school grades and the LSAT were entered together, the correlation between predicted and actual pass or fail was .52 with data pooled across jurisdictions and .58 when a model was built for each of 39 jurisdictions. A correlation of .52 corresponds to about 27 percent of the variance, our arithmetic, and the report states that a considerable amount of unexplained variance remains after both predictors are taken into account. In 87 percent of pairings of one examinee who passed with one who failed, the model gave the passer the higher predicted probability, and in 12.7 percent it ranked the pair the wrong way (0.3 percent were tied). Undergraduate GPA and the selectivity of the undergraduate school made no practical additional contribution once law school grades and the LSAT were in the model.

The later studies ask what law school adds. The AccessLex report found law school grades the strongest predictor even early in school. Pals and colleagues tested whether first-year grades sit between the admission credentials and the exam, and they summarize the earlier AccessLex result as an odds ratio for the LSAT that fell from 1.71 to 1.14 once first-year grades entered the model. The Texas Tech study found the same ordering in one school and added course-level detail. Grades in Civil Procedure accounted for 25.3 percent of the variance in bar score and grades in the first-year legal research and writing course for 17.8 percent, which the authors surmise reflects study skills, stamina and feedback as well as the skills taught, although they did not test that.

None of this contradicts the test maker. LSAC's validity summary reports a correlation of .40 between the LSAT and first-year law grades in an earlier, 2019 study, and .60 after adjustment for restriction of range, and the page on IQ and the LSAT covers LSAC's later study. It also says that third party research has found the LSAT more highly predictive of bar exam performance than any other factor available to law schools before matriculation, while cautioning against uses of the score other than holistic admission decisions. Both statements can hold at once. On the evidence reviewed here, the LSAT is the strongest of the credentials available before law school, and law school grades, which exist only later, are stronger still.

The pattern is the one Kuncel and Hezlett found across graduate and professional fields in their 2007 review in Science. Standardized admission tests predict later success, tests with content tied to a discipline tend to predict better than broad verbal and quantitative tests, and the most accurate prediction combines tests and grades. Law grades are in effect a discipline-specific measure, records of performance in the same doctrinal subjects that the bar later samples.

Three limits keep these results in proportion. The designs are observational, and the samples are students who finished law school and chose to sit, a selected group. Two of the four rows are AccessLex Institute reports, one labeled a working paper, and the Texas Tech study covers one school and one jurisdiction. And a predictor that explains a quarter or half of the variance in a score is useful for advising students and spotting those who may need support, not for ruling any individual in or out.

7 Where Does General Ability Enter the Picture?

General ability probably enters upstream of the exam, through who reaches law school, how fast the law is learned and the grades that result, but no study reviewed here measured it, and a bar score alone cannot separate it from knowledge, time and circumstance. General ability, g, is the variance shared across cognitive tasks of different content, and the page on the g factor explains how it is estimated. Its relationship with schooling is well documented. A meta-analysis by Roth and colleagues of 240 independent samples with 105,185 participants found a population correlation of .54 between intelligence tests and school grades, after correcting for sampling error, measurement error and range restriction. The page on IQ and academic achievement covers that evidence in full. Most people who sit the bar have completed a bachelor's degree, an admission test and law school, and stages that rely on grades and test scores plausibly select on abilities that correlate with g, although this review found no study that measured it.

How would g act on a knowledge exam? The best known account is a learning account. Kuncel and Hezlett's 2010 review says that ability tests predict performance because they forecast how much relevant knowledge and skill a person has now and will go on to acquire, and it cites research on job performance indicating that job knowledge and skills mediate the relationship between cognitive ability and job performance. The account explains why ability matters for knowledge-based outcomes. No study reviewed here tested it on the bar exam.

Applied to the bar, the account yields predictions that the studies in the previous section are consistent with. A person with stronger general ability should on average learn the doctrine faster and reason from rules to facts more accurately, and so earn higher law school grades. Because law grades sit closer to the exam in content than any ability test does, they should predict the exam better than an ability test would. And because everyone in the pool has already been selected on criteria linked to ability, the differences that remain among bar takers should owe more to knowledge, time and effort than differences in the general population do. This is our interpretation and not a finding, because none of the studies measured g.

Selection also hides correlation. LSAC's .40 and .60 for the LSAT and first-year grades differ because of restriction of range, which arises when a sample is selected on the predictor, as a law school class is selected on LSAT score. Frey and Detterman show the same effect in a general ability setting: the SAT correlated .483 with Raven's Advanced Progressive Matrices in 104 undergraduates, and .72 after correction for restricted range. A pool of bar takers has been selected at more stages than either group, so correlations computed inside it probably understate the relationship that would hold across all adults.

The missing piece is a joint study. This review found no published study in which the same people took a named cognitive battery and a bar exam. For the SAT such evidence exists. Frey and Detterman extracted g from the ASVAB for 917 participants in the National Longitudinal Survey of Youth 1979 and found a correlation of .82 with SAT scores, .86 corrected for nonlinearity, which is the kind of basis an SAT-to-IQ equation requires, and the page on IQ and the SAT covers the admission test for which such a joint sample exists. For the bar exam no comparable basis exists. A correlation of .80 between the LSAT and g circulates on web pages that rank for this topic. The searches for this page found no primary source for it, so the page does not use it.

8 How Much of Passing Is Preparation, Time and Persistence?

Bar results move substantially with preparation, time and persistence, which a bar score alone cannot separate from ability. One piece of evidence is what happens when the same people sit again. In a report on the 2,293 people who first took the Texas bar exam in July 2004, Klein and Bolus found that the 412 who took it at least twice raised their average total scaled score by 34.8 points between the first and second attempts. The authors describe that as two thirds of a standard deviation in the population of all takers and attribute it probably to additional study rather than simply practice, since candidates already know the question types. Repeaters are selected because they scored below the pass mark the first time, so some of the gain is expected from regression toward the mean, and the report does not separate that from the effect of study. Seventeen percent lost ground, and the correlation between first and second attempt totals was .54, which the authors read as meaning that a first score gives only a rough estimate of later standing. Two thirds of a standard deviation on the IQ scale would be 10 points, our arithmetic, and a retake is not read as a ten point change in intelligence. Whatever its exact cause, it is a change in performance on a content exam, not a measured change in ability.

Time and circumstances show in the AccessLex report. Graduates with below average first-semester grades whose grades did not grow before graduation had a 25 percent chance of passing, against 43 percent among similar peers whose grades grew by an average of about 0.17 grade points. Graduates who spent more than 21 hours a week on responsibilities such as caring for dependents or working a job unrelated to law had lower third-year grades and lower odds of passing than peers who spent 0 to 5 hours on them. Working in a law-related job, feeling that law school contributed very much to skills development and regular class participation were each modestly associated with passing. These are associations and not experiments, but they describe a process in which hours, support and feedback matter.

Admission tests move with coaching too, though less. Kuncel and Hezlett's 2007 review reports that the typical gain from coached preparation for standardized admission tests is about 25 percent of one standard deviation.

The contrast with an IQ score is the point. A bar score is meant to reflect how much of the required knowledge and skill a candidate has at the time of the sitting, and it is expected to change with study, which is why retakes are allowed and why NCBE tracks first-time and repeat takers separately. An IQ score is meant to describe standing among adults on broad abilities and not the content of one course of study, and the page on preparing for an IQ test separates preparation that reduces noise from practice that inflates a score. The page on whether IQ can be improved reviews which interventions change measured ability and which change only the practiced task. The two numbers answer different questions about change.

9 Why Do Pass Rates Differ by State, School and Sitting?

Pass rates differ by state, school and sitting because pass marks, candidate pools and repeat taker shares differ, so a rate describes a group at a sitting more than anyone's ability. NCBE's 2025 statistics snapshot, built from jurisdiction reports, counted 67,442 people taking a bar exam in 2025, of whom 63 percent passed. The 46,356 first-time takers passed at 76 percent and the 21,086 repeat takers at 34 percent. Repeaters made up 61 percent of the 19,645 February examinees, and first-time takers passed at 60 percent in February against 79 percent in July. Graduates of non-ABA-approved and foreign law schools were 4 and 15 percent of examinees and passed at 32 and 35 percent. Each of those splits changes who is in the pool, and the pool changes the rate.

State differences are large. In NCBE's table of results by jurisdiction, read on October 6, 2026, overall pass rates for July 2025 among the 50 states and the District of Columbia ran from 54 percent in Alabama to 86 percent in Utah, and first-time rates from 66 percent in Connecticut to 89 percent in Utah. NCBE warns that jurisdictions define first-time and repeat takers differently. The pass mark is only part of the story, because Alabama and Utah both use a legacy UBE minimum of 260 and sit at opposite ends of that range. The mix of candidates, the schools in and near the state and where graduates choose to sit all enter.

Schools differ too, though most of the variation sits within schools. The ABA reported on March 11, 2026 that first-time takers in 2025 achieved an aggregate pass rate of 84.10 percent and that 92.15 percent of 2023 law graduates who sat for a bar exam passed within two years. The first figure is higher than NCBE's 76 percent for all first-time takers in 2025, which is consistent with NCBE's pool including graduates of non-ABA and foreign schools, although the two sources do not use identical definitions. In the sample of 39 schools used by Pals and colleagues, the intraclass correlation for first-time passing was 0.12, meaning that 12 percent of the variation lay between schools and 88 percent within them. In the national study, the probability of passing still varied by group of law schools and by region after law school grades and the LSAT were taken into account.

10 What Does Failing the Bar Say About Ability?

Failing the bar says that on one sitting a candidate's performance fell below one jurisdiction's pass mark, which is a fact about a score and a standard and not a verdict on intelligence. Most people who fail do not stay failed. In the LSAC national study, 88.62 percent of the 23,103 graduates with results passed on the first attempt and 94.8 percent had passed eventually, and among those who failed first and later passed, nearly three quarters did so on the second attempt. In the Texas cohort, 79.3 percent passed at once and 92.2 percent within four attempts, and 294 of the 412 repeaters, 71 percent, passed.

In the Texas Tech study, failure clustered among lower law school grades: 39 alumni failed on a first attempt in July 2014, and by the paper's table one had graduated in the top quarter of the class, five in the second, 12 in the third and 21 in the fourth. That concentration fits the picture in the earlier section, where law school performance is the strongest predictor. It says nothing about cognitive ability apart from whatever the grades capture, and the grades capture effort, preparation and fit with the school's assessments as well as reasoning.

Circumstances matter as well. In July 2026, about 645 registered Washington applicants could not test because of a venue failure, an extreme case, and the AccessLex report links hours of caregiving and unrelated work to lower odds of passing. A score on a given day reflects knowledge, preparation, conditions and chance. An exam that measures acquired knowledge against a pass mark cannot tell a candidate who lacked ability from one who lacked time, which is part of why every study of predictors leaves so much unexplained.

What a failing score does not say is just as important. It is not a clinical statement, it does not estimate IQ, and it gives no basis for inferring low general ability. A passing score does not show high general ability either. The honest reading is conditional and narrow: on that sitting, against that jurisdiction's pass mark, the candidate's demonstrated knowledge and skills fell short, and in the cohorts reported here most candidates who tried again eventually passed.

11 Why Can No Bar Score Be Converted to an IQ?

No bar score converts to an IQ because the two numbers differ in construct and reference group, no study links them in the same people, and two correlations cannot be chained into a third. Each reason is independent, so the conversion fails even if one of them is set aside.

The construct differs. A bar exam samples acquired legal knowledge and lawyering skills, and an IQ estimates general ability and broad abilities among adults. The reference group differs. An IQ places a person among adults of their age, and a bar score compares a candidate with a standard held constant across sittings, while any percentile attached to it is a percentile inside the pool of people who sat. The table shows what percentile matching would produce if a UBE score were put on the IQ scale through its percentile inside a pool. It uses NCBE's published 2025 means and standard deviations and assumes a normal distribution, so it is our arithmetic and an illustration only.

Read the last column as a demonstration, not a result. It treats the average person in each pool as an IQ of exactly 100 and assumes a normal distribution, and neither assumption is supported for people who sit the bar. The same 266, the pass mark in 13 jurisdictions, lands at the 28th percentile of the July pool and the 58th of the February pool, so percentile matching would put it about 12 points apart on the IQ scale. The gap comes from nothing but which sitting serves as the reference group, and neither value describes anyone's standing among adults. A pool of people who have completed a law degree is not a sample of adults, and a percentile earned inside it answers a different question from the percentile that defines an IQ. The page on converting a percentile to an IQ shows how the conversion works when the percentile does come from the general population, and the page on IQ score versus percentile explains why the two scales cannot be swapped without that.

Correlations cannot be chained either. Two correlations leave the third within a wide band: hypothetical correlations of .50 and .70 allow a third anywhere from about -0.27 to 0.97, our arithmetic, so a published link between the bar exam and the LSAT and another between the LSAT and g, even if both existed, would not give a link between the bar exam and g. The bar exam is also used as a benchmark for artificial intelligence systems, a different question that the page on ChatGPT and IQ covers.

What to use instead depends on the question. If it is where fluid reasoning (Gf), crystallized knowledge (Gc), working memory (Gwm) and processing speed (Gs) stand among adults, a normed cognitive battery answers it directly. If it is whether a candidate will pass the bar, the better evidence is law school grades and the pass mark of the state. If it is how rare an IQ is, the page on how rare an IQ is gives it for an IQ, not for a bar score.

12 What the Evidence Supports, Stated Narrowly, and How to Read a Score

The evidence supports a narrow set of statements about the bar exam and ability, and the Standards for Educational and Psychological Testing say how far any score should be pushed beyond them. The Standards for Educational and Psychological Testing, published jointly by AERA, APA and NCME in 2014, ask in Standard 11.13 that the content domain of a credentialing test be defined and justified by its importance for credential-worthy performance, and in Standard 11.16 that the pass level rest on that performance and not be adjusted to control how many people pass. APA's testing and assessment page presents the Standards as a product of the three organizations since 1966, makes them open access, and records that a joint committee was named in June 2024 to revise the 2014 edition. The rule that follows is simple. Read a score against the purpose, population, scale and decision it was built for.

Stated narrowly, the evidence on this page supports five things. The bar exam measures acquired legal knowledge and lawyering skills against a pass mark that each jurisdiction sets. Law school grades and the LSAT both predict passing, grades more strongly once a student is enrolled, with a correlation of about .5 in the one national model and much left unexplained. General ability probably contributes through selection and the speed of learning, but no study reviewed here measured it directly. Preparation, time and persistence move the score substantially, and most candidates who fail eventually pass. And no bar score converts to an IQ.

Several things the topic does not say about the individual reader matter as much. Passing says nothing certain about IQ, and neither does failing. Group findings are not predictions for any person. Nothing here is clinical. A bar result is also not a measure of how good a lawyer someone will be, because the Standards make the point that credentialing exams are not meant to predict individual performance in a specific job. The page on average IQ by profession covers occupational ability estimates, and the page on IQ and job performance covers what ability does predict at work.

We sell a paid, online, self-administered cognitive assessment, so treat this paragraph as a disclosure and check it against the rest of the page. ACIS reports six indices and a Full Scale IQ on the standard scale, with a mean of 100 and a standard deviation of 15, plus percentiles and a 95 percent confidence interval, against adult norms for ages 16 to 90. On October 6, 2026, the Quick form cost 15 dollars for 6 subtests and about 45 minutes, the Optimized form 30 dollars for 13 subtests and about 110 minutes, and the Full Scale form 50 dollars for all 20 subtests and about 175 minutes, each as a one time payment with no subscription, breaks allowed and 30 days to finish. There is a free trial that needs no card and a 5 day quality guarantee, and prices can change. The limits are plain. ACIS is 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 available in English only. Nothing on this page is evidence that its scores predict bar results, because none of the studies cited used it. The technical manual documents the instrument, and the page on professional IQ tests explains the routes for people who need a signed report from a licensed examiner.

What a reader should do follows from the question. A candidate asking whether they will pass should look at law school grades, the pass mark of the jurisdiction and NCBE's official materials, not at an IQ. Someone asking where their fluid reasoning (Gf), crystallized knowledge (Gc), working memory (Gwm) and processing speed (Gs) stand among adults can use a normed battery such as a Full Scale IQ test, and should not use that result to answer the first question or the bar result to answer the second.

13 Sources Behind This Page

Every figure on this page is traceable to one of the following sources, and each is linked at the point where it is used. Web pages were read on October 6, 2026 and can change. The calculations labeled as our arithmetic in the body are computed on published numbers and are not findings of the sources: the average time per MBE question, the 12 hours of testing, the share of variance behind a correlation, the 0.29 standard deviation between pass marks, the percentiles and IQ-scale values in the conversion table, the 10 point equivalent of two thirds of a standard deviation, and the range allowed for a third correlation in one hypothetical example. The 2021 AccessLex report is labeled a working paper, and the 2024 AccessLex paper was read as published on the institute site.

• National Conference of Bar Examiners. Pages on the Uniform Bar Examination, the Multistate Bar Examination, the Multistate Essay Examination, the Multistate Performance Test, minimum UBE scores, the NextGen UBE (about, content, scores, sample questions) and bar exam results by jurisdiction, with the press releases of July 31, 2026 and September 25, 2026. ncbex.org, read October 6, 2026.
• National Conference of Bar Examiners. 2025 Statistics: the Multistate Bar Examination, the Uniform Bar Examination and the 2025 Statistics Snapshot. The Bar Examiner, Spring 2026, read October 6, 2026.
• Case S M, in The Testing Column: Scaling, Revisited, which reprints her articles of 2005 and 2006 on scaling written scores to the MBE. The Bar Examiner, Fall 2020, volume 89, issue 1, pages 68 to 75.
• NCBE Testing Task Force. Final Report of the Testing Task Force. National Conference of Bar Examiners, April 2021.
• Wightman L F. LSAC National Longitudinal Bar Passage Study. Law School Admission Council, LSAC Research Report Series, 1998.
• Klein S P and Bolus R. Initial and Eventual Passing Rates of July 2004 First Timers. GANSK and Associates for the Texas Board of Law Examiners, June 9, 2006.
• Austin K A, Christopher C M and Dickerson D. Will I Pass the Bar Exam: Predicting Student Success Using LSAT Scores and Law School Performance. Hofstra Law Review, 2017, volume 45, issue 3, pages 753 to 784.
• Taylor A N, Scott J M and Jackson J. Predicting Law School and Bar Success, the national report of findings for the AccessLex and LSSSE Bar Exam Success Initiative. AccessLex Institute, working paper last updated March 24, 2021, read October 6, 2026.
• Pals A M, Scott J M and Wilson P. Predicting Bar Success: The Mediating Effects of Law School GPA. AccessLex Institute, 2024.
• Camara W J. Evidence to Support Validity Claims for Using LSAT Scores in Law School Admission. Law School Admission Council, read October 6, 2026.
• American Bar Association, Council of the Section of Legal Education and Admissions to the Bar. Bar exam pass rates increased in 2025, including for first-time exam takers. ABA news release, March 11, 2026.
• Kuncel N R and Hezlett S A. Standardized Tests Predict Graduate Students' Success. Science, 2007, volume 315, issue 5815, pages 1080 to 1081.
• Kuncel N R and Hezlett S A. Fact and Fiction in Cognitive Ability Testing for Admissions and Hiring Decisions. Current Directions in Psychological Science, 2010, volume 19, issue 6, pages 339 to 345.
• Roth B, Becker N, Romeyke S, Schäfer S, Domnick F and Spinath F. Intelligence and school grades: A meta-analysis. Intelligence, 2015, volume 53, pages 118 to 137.
• Frey M C and Detterman D K. Scholastic Assessment or g? Psychological Science, 2004, volume 15, issue 6, pages 373 to 378.
• American Educational Research Association, American Psychological Association and National Council on Measurement in Education. Standards for Educational and Psychological Testing. AERA, open access edition, 2014, chapter 11, pages 169 to 182; see also the APA page on the Standards, read October 6, 2026.

14 Frequently Asked Questions