IQ vs LSAT: Related, Not Equal

0 The Short Answer

The LSAT is the admission test that comes closest to a pure reasoning exam, and it still cannot produce an IQ score. There is no algebra on it, no vocabulary list to memorize, no science passage that rewards having taken chemistry, no formula sheet. Every scored question hands you the material you need inside the question itself and then asks what follows from it. That design is why the comparison with cognitive testing keeps coming up, and why it deserves a more careful answer than the usual one.

The careful answer has three parts. First, the LSAT compares you against people who decided to apply to law school, a group that has already survived four years of undergraduate selection, not against adults your age in general. Second, the 120 to 180 scale is an equating scale built so that scores mean the same thing across test forms, and it carries no relationship at all to a mean of 100 and a standard deviation of 15. Third, the current test samples exactly two channels of ability, formal argument analysis and dense reading, and says nothing about the other things a cognitive battery measures separately.

So a high LSAT score is real evidence of something real. It is evidence that you can hold a chain of premises in mind, spot the gap between what an argument asserts and what it needs, and do it against a clock, in English, on prose written to be slightly harder than it needs to be. That is a genuine cognitive accomplishment. It is also a narrower statement than "this person has an IQ of X," and the rest of this page is about exactly how much narrower.

1 What the Test Looks Like After August 2024

Most of the LSAT commentary written before 2024 describes a test that no longer exists, so start with the current shape. LSAC's official FAQ states that "Starting with the August 2024 LSAT, the multiple-choice portion of the test consists of two scored Logical Reasoning sections and one scored Reading Comprehension section." A fourth section, either Logical Reasoning or Reading Comprehension, is administered unscored so that LSAC can pretest questions for future forms. You are not told which one it is, so you have to treat all four as real.

The timing is unchanged: four sections of 35 minutes each, a 10 minute intermission between the second and third, and roughly three hours for a standard administration. LSAT Argumentative Writing is a separate proctored task that candidates complete on demand, and it does not feed into the 120 to 180 score at all. Law schools can read it; the number on the score report does not contain it.

The removed piece is the one everyone remembers. Analytical Reasoning, universally called logic games, gave you a scenario with a set of constraints and asked which arrangements were possible. It was the section people diagrammed, and diagramming is the reason it is gone: LSAC agreed to phase it out under a 2019 settlement brought by two blind test takers who argued under the Americans with Disabilities Act that a section whose standard solution method is drawing pictures put them at a structural disadvantage. It ran for the last time in June 2024 and was replaced by the second scored Logical Reasoning section.

That change matters more for this comparison than it does for admissions. Logic games was the most Gf loaded thing on the exam, the piece that most resembled a formal reasoning task with no verbal dressing, and it was also the piece most people could train into the ground. Removing it made the LSAT more verbally weighted than it has ever been, which pulls it further from a matrix reasoning test and closer to a very demanding reading and argument exam. Any older article that treats logic games as evidence that the LSAT measures raw reasoning is now describing a test that retired.

One more current fact worth knowing before anyone builds a theory about score meaning: LSAC allows five administrations within the current reportable score period and seven over a lifetime. Scores you see quoted are frequently not first attempts, which is a problem for the conversion tables discussed later.

2 What Each Section Actually Asks You to Do

Generic descriptions of the LSAT as "logic and reading" hide the interesting part, which is that the two scored section types make quite different cognitive demands.

Logical Reasoning presents a short argument, usually four to six sentences, followed by one question about it. The question types are stable and named: strengthen, weaken, necessary assumption, sufficient assumption, flaw, parallel reasoning, inference, principle, paradox. The core operation is separating what an argument says from what it needs in order to work. A stimulus claims that a policy reduced accidents because accidents fell after it passed; the task is recognizing that the argument silently assumes nothing else changed. This is formal inference wearing everyday clothes, and it is the closest thing on a mainstream admission test to a deductive reasoning task.

Reading Comprehension gives four passage sets, one of which is a comparative pair of shorter texts, drawn from law, humanities, social science and natural science. The passages are not selected to be interesting; they are selected to be structurally dense. Questions ask for main point, author attitude, the function of a specific paragraph, what can be inferred, and how the author would respond to a claim never stated in the passage. The demand is not vocabulary. It is holding an argumentative structure in memory while reading, under a clock that gives you roughly eight and a half minutes per passage set including the questions.

The retired Analytical Reasoning section was different from both. It was a constraint satisfaction task, closer to a puzzle than to prose, and it was the only part of the exam with a meaningful visual spatial component, since most successful test takers solved it by building a diagram. Its removal is the single clearest reason to treat pre-2024 claims about what the LSAT measures as historical.

Notice what is absent from that inventory. No arithmetic, no geometry, no data interpretation, no symbol substitution under time pressure, no memory span task, no figure rotation. If you want the fuller map of what a cognitive battery covers and why the missing pieces are not optional, what IQ tests measure lays out the domain structure that the LSAT samples one corner of.

3 How the 120 to 180 Scale Is Built

Every LSAT statistic people quote runs through this machinery, so it is worth understanding before interpreting anything.

Your raw score is the count of questions you answered correctly. LSAC's scoring page is explicit that "There is no deduction for incorrect answers," and that "The total number of questions you get right is what matters for your score, not which particular questions you get right or wrong." A hard question and an easy question contribute the same single point. Guessing is strictly rational on every item you cannot solve, which already distinguishes the LSAT from instruments where response patterns carry information.

That raw count is then converted to the 120 to 180 scale through an equating process, so that a given scaled score represents the same level of performance whether you happened to sit a slightly harder or slightly easier form. This is why raw score conversion charts found online are form specific and cannot be generalized: the number of questions you can miss and still reach 170 differs between administrations by design.

Two further pieces appear on the score report. The percentile rank tells you "the percentage of test takers whose scores were lower than yours during the previous three testing years," which means the denominator moves as the applicant pool moves. And the score band communicates that the reported number is an estimate with error around it rather than a fixed personal constant.

How much error? LSAC's reliability and validity page reports that LSAT forms "typically have reliability coefficients over 0.9," and that mean reliability across the eight administrations in the 2021 to 2022 testing year was above 0.90 for every one. That is high, comparable to a well built cognitive battery, and it is exactly why the score band exists rather than being hidden. A reliability of .90 is not a promise that a repeat sitting returns the same integer. It is a statement that most of the variance in scores is signal, with a measurable slice that is not.

4 The Norm Group Is Law School Applicants, and That Changes Everything

This is the single most important fact for anyone tempted by a conversion table, and it is the fact those tables always skip. A score means something different when the comparison group is adults in general, which is the premise of taking an IQ test as an adult rather than of reading a rank built from people who already selected themselves into a profession.

An IQ score is anchored to a general population sample stratified by age. The distribution is engineered so that 100 is the median of adults in general and each 15 points is one standard deviation. An LSAT percentile is anchored to people who registered for the LSAT in the last three testing years. To sit in that group you have to want to be a lawyer, which in practice means you have almost certainly finished or nearly finished an undergraduate degree, you have self selected into a verbally demanding profession, and you have paid a nontrivial registration fee to be there.

The consequence is that LSAT percentiles and IQ percentiles are not translatable even in principle, because they answer different questions. A 60th percentile LSAT means "better than 60 percent of aspiring lawyers." A 60th percentile IQ means "better than 60 percent of adults." Those two sentences describe different populations, and the second group is much broader than the first at every level of the distribution. The general logic of reference groups is worked through in more depth in the IQ percentile chart, which is worth reading before comparing any two percentile figures from different tests.

LSAC's own data shows how compressed the enrolled end of this pool is. In the Summary of 2021 to 2025 LSAT Correlation Study Results, published by LSAC in February 2026, the roughly 170 participating law schools had a mean entering class LSAT of 156.54 in the 2025 study, with school means running from 144.76 at the low end to 169.59 at the high end, and the middle half of schools between 152.62 and 160.43. The entire American legal education system, from the least selective accredited school to the most, fits inside about 25 points of a 61 point scale.

That compression is the practical reason percentile matching produces nonsense. Small movements on the LSAT scale correspond to large movements in applicant rank, because almost everyone taking the test is clustered in the middle of it. Mapping that steep local slope onto a general population bell curve manufactures spectacular IQ estimates out of ordinary score differences.

5 "No Content Knowledge" Is Not the Same as "No Background Advantage"

The LSAT's reputation as a reasoning test rests on a true observation: it tests no legal doctrine, no history, no formulas. LSAC's position is that law faculty rate the assessed skills as those most essential to law school and legal practice, not that candidates should arrive knowing law. Nothing on the exam requires facts you were supposed to have learned. Tests that try to strip background out rather than merely omit content hit a version of the same wall: the classic culture-fair instruments, Raven's Progressive Matrices and the Cattell CFIT, are built entirely from abstract figures and still only reduce cultural and educational loading, which is why culture-reduced is the accurate label.

It does not follow that background is irrelevant. Three kinds of prior learning are load bearing on the current test, and none of them is content knowledge in the usual sense.

• Academic reading fluency. The passages are written in the register of an academic journal. Someone who has spent four years reading arguments in that register decodes them faster, and speed is scored. This is acquired skill sitting inside what looks like a pure reasoning measure.
• The conventions of formal argument. Necessary versus sufficient conditions, the difference between correlation and cause, what a counterexample does to a universal claim. A philosophy graduate has met all of this explicitly. Someone who has not can learn it in a few weeks, but they start behind, and the starting gap is not a difference in reasoning capacity.
• English as a first academic language. Reading Comprehension is timed tightly enough that a small decoding disadvantage compounds across four passage sets. The section is measuring reasoning through a language channel, and the channel is not neutral.

In the vocabulary of the Cattell Horn Carroll framework, this is the difference between fluid reasoning and crystallized comprehension knowledge, and the LSAT loads on both at once rather than isolating either. The distinction and why it matters for score interpretation is covered in fluid versus crystallized intelligence. A test can be free of memorized content and still reward a decade of accumulated verbal exposure, which is precisely the LSAT's situation.

6 Formal Logic and Fluid Reasoning: The Real Relationship

Now the question this page exists for. Logical Reasoning asks you to determine what must follow from stated premises. A matrix reasoning item asks you to determine what must come next in a pattern of figures. Both feel like "reasoning." Are they the same thing?

They are related and not identical, and the psychometric literature has a precise way of saying so. In John Carroll's 1993 survey of the factor analytic evidence, Human Cognitive Abilities, fluid reasoning is not a single narrow ability but a cluster of them, of which two are relevant here. Induction is the ability to discover the rule that governs a set of instances, which is exactly what matrix items require, since nobody tells you what the pattern is. General sequential reasoning is the ability to start from stated premises and rules and derive a conclusion by valid steps, which is exactly what Logical Reasoning requires, since the premises are handed to you and the rule of inference is the whole task. Both narrow abilities sit under the same broad factor, which is why an instrument aimed at fluid reasoning samples several task families instead of matrices alone.

Those two narrow abilities sit under the same broad factor and correlate substantially, which is why a strong matrix performer usually finds LSAT argument structure tractable and vice versa. They are not interchangeable, though. Induction is generative: you must invent a candidate rule and test it. Deduction is evaluative: the rule is present and you must apply it without leaking assumptions. People differ in which of the two they find easier, and the difference shows up as exactly the sort of profile unevenness a single composite hides.

There is a second complication, and it is bigger. On the LSAT, that deductive operation is delivered entirely through English prose about real world topics, which means performance depends on comprehension of the stimulus as much as on inference from it. A matrix item deliberately strips language out so that comprehension cannot contaminate the reasoning estimate. So the honest statement is that LSAT Logical Reasoning measures deductive reasoning through a verbal channel, and a matrix task measures inductive reasoning through a nonverbal one. They overlap in the reasoning term and diverge in everything else.

7 The Four Domains the LSAT Never Touches

A cognitive battery reports separate indices because performance across domains is genuinely uneven within the same person. The current LSAT collapses everything into one number drawn from two verbal channels, which leaves four broad domains completely unsampled. An instrument built the opposite way carries the mirror image blind spot, because a nonverbal battery delivered through figures and matrices concentrates on fluid reasoning and visual spatial processing and never observes the verbal argument evaluation this exam is made of.

Visual spatial processing disappeared from the exam when logic games did. Nothing on the current test asks you to rotate, fold, assemble or track anything in space. Someone whose strongest domain is visual spatial reasoning receives no credit for it and no evidence about it.

Quantitative reasoning is absent by design. The LSAT is unusual among admission tests precisely here: the GRE and the GMAT both carry substantial quantitative sections, so a candidate strong in quantitative reasoning and average in verbal reasoning posts a very different profile on those exams than on this one. The LSAT gives that person nowhere to show it.

Working memory is used constantly but never isolated. Holding a four premise argument while scanning five answer choices clearly loads memory, yet the score cannot separate a memory limitation from a comprehension limitation from a reasoning limitation. When those three are confounded, a low score tells you something went wrong without telling you what.

Processing speed is measured only as a confound. The 35 minute limit means speed contributes to every scored item, but there is no untimed comparison, so a slow and accurate reasoner and a fast and sloppy one can land on the same scaled score for opposite reasons. A battery that reports speed separately can tell those two people apart. The LSAT, by design, does not need to.

None of this is a defect. LSAC built an instrument to predict first year law school grades, and grades in law school are earned largely by reading dense text and constructing arguments about it under deadline. The exam samples the thing it needs to sample. The error is only in reading a purpose built admission measure as though it were a broad description of a mind.

8 Trainability: The LSAT's Best Feature and Its Worst One for This Comparison

Ask anyone who has prepared seriously and they will tell you scores move. The question is what moves with them. Employers face the same question on a twelve minute scale, and the answer there is well documented: practice on a Wonderlic or CCAT buys format familiarity and speed, not the underlying ability the screener is trying to sample.

Start with LSAC's published behavior data. In technical report TR 26-01, covering the 2018 to 2019 through 2024 to 2025 testing years, LSAC reports that 49.1 percent of all LSAT takers in the 2024 to 2025 testing year tested more than once, and that first time takers scored around 1.5 points lower on average than repeat test takers. Read that second figure carefully, because it is a comparison between two groups rather than a measured within person gain, and repeaters are self selected for being motivated and for having been disappointed the first time. What it establishes is not the size of the practice effect. It establishes that roughly half the reported scores in circulation are not first attempts.

Now the mechanism. Preparation for this exam works because the question types are finite and named. Learning to recognize a necessary assumption question on sight, learning that "the argument fails to consider" signals a specific answer shape, learning to spend eight minutes rather than eleven on a passage set: these are strategy acquisitions that convert directly into scored points. That is a real skill gain and it is exactly what the test intends to reward, since law school also rewards learning the conventions of a genre quickly.

What it is not is evidence of a change in general reasoning capacity. The relevant literature is on transfer, and it is discouraging in a consistent direction. Sala and Gobet's 2017 review in Current Directions in Psychological Science, which examined chess instruction, music instruction and working memory training together, found that far transfer to general cognitive ability rarely occurs, and that apparent effects shrink as study designs improve and add active control groups. Training on a task reliably improves that task and things resembling it, and reliably fails to improve distant abilities.

Put those two facts side by side and the conversion problem becomes obvious. A score that moves 8 points with three months of structured practice and a score built to estimate a stable trait cannot be two expressions of the same quantity. If they were, the same three months would have moved measured cognitive ability by half a standard deviation, which nothing in the training literature supports.

9 What LSAC's Own Validity Numbers Actually Say

The LSAT has better published validity evidence than almost any consumer facing test, and the numbers are more interesting than either its defenders or its critics usually admit. All figures below come from LSAC's Summary of 2021 to 2025 LSAT Correlation Study Results, dated February 2026, covering roughly 170 participating law schools per year. Publishing correlations that critics can check is what a serious instrument does, and evidence of that kind belongs with the reliability, validity and norms behind an accurate score.

Three things in that table deserve comment. The first is the size of the raw correlation. A median of 0.42 against first year grades is strong for an admission test and weak as a description of a person: it leaves the large majority of grade variance unexplained, which is why LSAC reports score bands and why schools are told to use the score alongside other evidence.

The second is the gap between raw and corrected values. The jump from 0.42 to 0.62 is entirely a statistical correction for the fact that law schools only observe grades for students they admitted, which truncates the range of scores in the sample. That correction is legitimate and standard, and it is also a reminder that every correlation computed inside a selected group understates the relationship that exists in the wider pool. Any table converting LSAT to IQ using data from admitted students inherits this problem and never mentions it.

The third is the strangest number in the report and the most instructive. Among matriculants, the raw correlation between LSAT and undergraduate GPA is slightly negative, around minus 0.05, and turns positive at about 0.25 only after correcting for restriction of range. In the general applicant population, higher LSAT scores go with better grades. Inside an admitted class, they trade off against each other, because admissions offices accept a strong score with weaker grades and vice versa. That is selection creating a false negative relationship out of a genuinely positive one, and it is the cleanest illustration available of why correlations from selected samples cannot be read as facts about people.

10 Why LSAT to IQ Conversion Tables Have No Legitimate Basis

Search for a conversion and you will find charts asserting that a 170 corresponds to an IQ around 145 and that a 180 lands somewhere near 160. These are not disputed estimates. They are constructions with no data behind them, and it is worth naming the four separate failures rather than waving at them. The argument is identical for the GRE: a selected norm group, trainable content, and a scale built for a different job.

There is no joint sample. A defensible conversion requires a group of people who took both an LSAT and a named cognitive battery within a reasonable interval, with both scores recorded. No such dataset is published. Every chart in circulation is built by matching percentiles across two distributions that were never observed in the same people, which is arithmetic rather than measurement. The same arithmetic fails one step earlier for an instrument with no reference sample at all: the public domain ICAR item pool publishes reliability and validity evidence but never published a table turning a raw score into a percentile, so there is nothing legitimate to convert from.

The reference populations are incompatible. Percentile matching only works if both percentiles describe the same population. LSAT percentiles describe recent LSAT registrants; IQ percentiles describe adults in general. Equating the 99th percentile of a self selected graduate pool with the 99th percentile of the population is not a conservative approximation. It is a category error that inflates in a predictable direction, always upward.

The output implies rarities nobody is claiming. On the modern scale, 130 is roughly the top 2 percent, about 1 person in 44. A score of 145 is around 1 in 740 and 160 is around 1 in 31,000. Tens of thousands of people have scored 170 or above on the LSAT over the past decade. A conversion that assigns all of them a 1 in 740 rarity is asserting something arithmetically impossible about the size of the applicant pool. If you want the underlying rarity figures laid out properly, what counts as a good IQ score covers the ladder.

The input is not a stable quantity. With roughly half of test takers sitting the exam more than once and reported scores commonly reflecting a prepared and repeated attempt, the number fed into the conversion has already absorbed practice effects that a cognitive score is specifically designed to resist. Converting a trained score into a trait estimate imports the training as if it were ability.

The honest version of what a high LSAT supports is a conditional statement, not an equivalence: people who score 170 or above are, as a group, very likely to be well above average on verbal reasoning and reading comprehension measures. Group likelihood is not individual measurement, and a conditional statement cannot be inverted into a number for one person.

11 What a 170 Licenses You to Say, and What It Does Not

Take a concrete case, because the abstraction gets slippery. Someone reports a 172, achieved on a second sitting after four months of preparation. What has been established?

Established with confidence: this person can evaluate written arguments accurately and fast, sustain attention across three hours of dense material, extract structure from academic prose, and learn a formal genre well enough to perform near its ceiling. Given LSAC's reliability figures, a repeat sitting would very probably land close to that number. Given the correlation study, this person is a strong bet for first year law school performance relative to classmates.

Established with moderate confidence: verbal reasoning and comprehension knowledge are likely well above the general adult average, because the floor for competitive performance on this exam is genuinely high and the test is not gameable without real comprehension.

Not established at all: quantitative reasoning, visual spatial processing, working memory span, processing speed, and therefore any full scale composite, since a composite is the aggregation of domains that were never measured here. Also unestablished is how much of the four month gain reflects strategy rather than capacity, because the score cannot separate them.

The asymmetry runs in one direction and it is worth stating plainly. A very high LSAT is meaningful positive evidence about part of a cognitive profile. A mediocre LSAT is much weaker negative evidence, because so many non ability factors can depress it: unfamiliarity with the format, English as a second academic language, a bad testing day, insufficient preparation time, or simply reading at a pace that a 35 minute section punishes. High scores are hard to achieve accidentally. Low scores are easy to achieve for reasons that have nothing to do with reasoning capacity. Intuition does not close the gap either, since self estimated ability correlates about .33 with measured ability in the Freund and Kasten meta-analysis, which leaves a candidate's own reading of a 172 weaker evidence than the domains the exam never sampled.

That asymmetry, incidentally, is also why the LSAT works for its actual job while failing at this one. Admissions offices only need the top of the distribution to be informative. A profile description needs the whole range to mean something.

12 If the Question Is the Profile, Measure the Profile

Most people who search for an LSAT to IQ conversion are not really asking about the conversion. They are asking one of two other questions, and both have better answers than a fabricated chart. One of those questions is simply what a broad cognitive profile would show, and what a good online IQ test should include starts with real norms, wide sampling of ability and an honest statement of limits.

The first is "am I smart enough for law school." The LSAT already answers that better than any cognitive test could, because it was validated against the exact outcome in question. LSAC's correlation studies exist precisely to establish that link, and no IQ battery has comparable evidence for predicting first year law grades. Trading a validated predictor for an unvalidated one is a downgrade, however impressive the second number looks.

The second is "what does my mind actually look like across domains," and here the LSAT genuinely cannot help, because it measures two channels and reports one number. If that is the question, the instrument has to sample the domains separately and report them separately, with intervals around each rather than a single point estimate. The relationship between cognitive scores and real world outcomes, which is what most people are actually curious about, is treated on its own terms in IQ and job performance.

ACIS is an online self assessment rather than a clinical evaluation, and it does not produce a credential any institution accepts. What it does is report performance across separate cognitive domains with confidence intervals, so that an uneven profile shows up as uneven instead of being averaged into a single figure. For a reader who just finished the LSAT, the interesting part is usually the contrast: the domains the exam never sampled are the ones most likely to surprise you.

Whatever you do with the score, resist the last temptation, which is treating either number as a verdict. A 156 is the average of every entering law class in the country and describes a competent reasoner. The people who most want a conversion table are usually the ones who would benefit least from believing one.

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.

• Law School Admission Council. The official LSAT pages: structure, scoring and administration.
• 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.