Joe Biden's IQ and the record that actually exists

0 Quick Answer

No intelligence test result for Joe Biden has ever been published, and there is no evidence that he has ever taken one. That is the entire answer to the search query, and every page that supplies a figure instead is supplying something else: a guess, a fabrication, or a score from an instrument that does not measure intelligence. ACIS has not assessed him, and no test can be administered to a person retrospectively.

What does exist is unusually rich for a page of this kind. In September 1987, during his first presidential campaign, Biden released his full academic file from the Syracuse University College of Law. Contemporaneous reporting in The Christian Science Monitor on 23 September 1987 recorded the contents: he ranked 80th in a class of 100 in his first year, 79th of 87 in his second, and 76th of 85 in his final year, and at the University of Delaware he had graduated 506th in a class of 688 with a C average. Those are documented facts about grades. They are not test scores, and the difference is the substance of this page.

The second document people reach for is a cognitive test given in presidential physicals. That instrument is the Montreal Cognitive Assessment, a 30 point dementia screen published by Nasreddine and colleagues in the Journal of the American Geriatrics Society in 2005. It takes about ten minutes, it was validated against mild cognitive impairment and mild Alzheimer's disease, and a perfect score on it carries no information about where a person sits on an intelligence distribution. Nothing on this page evaluates any officeholder's fitness, politics, or current condition. It evaluates documents.

1 The 65 Page File He Released in 1987

The reason this page has material at all is that Biden did something almost no public figure does: he released his complete academic record, including the parts that embarrassed him. The release happened in September 1987, in the middle of a presidential campaign, under pressure from reporting about claims he had made about his schooling. The file ran to 65 pages and covered his years at the Syracuse University College of Law.

The contents were reported contemporaneously and have never been retracted. The Christian Science Monitor account of 23 September 1987 lists the class ranks year by year: 80th in a class of 100 in the first year, 79th of 87 in the second, and 76th of 85 in the third. The same report converts those to standing, placing him in the bottom 20 percent in the first year, the bottom 10 percent in the second, and the bottom 11 percent in the third. It also records the undergraduate figure, 506th in a class of 688 at the University of Delaware, with a C average.

Two further items in the file matter for accuracy. The first is the scholarship claim. Biden had said he attended law school on a full academic scholarship. The released record showed the award covered roughly half of the 1,620 dollar tuition and was granted on financial need rather than academic promise. The second is the course grade discussed in the next section but one, which is the law review episode.

The class rank figure has been checked since. Snopes examined the specific claim that Biden finished 76th in a class of 85 at Syracuse and rated it true in a fact check dated 27 April 2024, syndicated here, noting that the figure was confirmed in decades old reporting by the Associated Press, The New York Times, and The Washington Post, and that the underlying documents were the ones Biden himself released. That is the rare case in this genre where a number attached to a public figure can be traced all the way to a document the person handed over voluntarily.

2 Where the IQ Question Came From in the First Place

Biden is one of the few American politicians whose name is attached to an IQ query for a specific and traceable reason: he raised the subject himself, on camera, in 1987. On 3 April 1987, at a campaign stop in Claremont, New Hampshire, a voter asked him about his academic credentials. The exchange was recorded and the footage survives in the C-SPAN archive.

In the recording he says he probably has a much higher IQ than the questioner does, and then makes a series of claims about his record: that he went to law school on a full academic scholarship, that he was the only person in his class with one, that after a poor start he finished in the top half of his class, and that he was the outstanding student in the political science department. He closes by offering to sit down and compare IQs.

Every claim in that passage was checkable against documents, and five months later the documents arrived. The scholarship was partial and need based. The finish was 76th of 85, which is the bottom eighth rather than the top half. The undergraduate standing was 506th of 688. The comparison of IQs was never possible in either direction, because neither man's score existed then and Biden's still does not. The offer was rhetorical, and it is the origin of the search query this page answers.

Two things follow from that, and they point in different directions. The first is that the phrase seeded a question with no answer, so anyone searching for the number is searching for something that was never produced. The second is that a claim about one's own standing is among the weakest kinds of evidence there is, whoever makes it. Pooled across 93 studies and 36,833 participants, self estimated intelligence correlates about .30 with measured ability, which is loose enough that a self assessment alone spans a very wide band. That arithmetic is worked through on the page on estimating your own IQ, and it applies to a senator in New Hampshire exactly as it applies to anyone else.

3 The 187 Figure and Where It Came From

The specific number circulating for Biden is 187, and it was fabricated. It is worth stating once, sourcing once, and correcting once, because the interesting question is not whether it is true but what kind of thing it is.

The claim appeared on 29 January 2024 in a post asserting that White House Press Secretary Karine Jean-Pierre had announced at that day's briefing that the president had taken an aptitude test showing an IQ of 187. Snopes examined it and rated it false in a fact check dated 30 January 2024. The fact check notes that both the video and the official transcript of the 29 January briefing show no such statement was made. It adds a point that gets lost in most coverage: an aptitude test is a different instrument from an intelligence test and does not produce an IQ score at all. The fact check concludes that it is unclear whether Biden has ever taken an intelligence test.

Consider what 187 would mean if it had come from an instrument. On the standard adult metric of mean 100 and standard deviation 15, 187 sits about 5.8 standard deviations above the mean. Under normal curve arithmetic that is a rarity on the order of one in tens of millions, and it is well beyond the range where any major adult battery reports a score at all, because the norm samples contain nobody at that position to anchor a conversion. A number that far out is not a strong claim. It is a claim the measurement system cannot express. You can check the arithmetic for any figure you like with the one in X rarity calculator.

The pattern is not specific to this figure or this person. Numbers attached to famous names almost never trace back to an administered test, and the ones that do are usually childhood ratio scores from a scale that was retired. The general tally is kept on the page on record high IQ claims, and the same trace has been run one name at a time on the celebrity figure audit.

4 Why a Class Rank Is Not an IQ Score

Class rank and measured ability are related, weakly, in a way that gets weaker the more selective the group. This is the single most useful thing to understand about the 1987 file, and it cuts in both directions: the rank does not establish a low score and a higher rank would not have established a high one.

Grades are a compound. They register how much a person knew, how much they worked, how well the assessment format suited them, how the curve was set, how much they attended, and what else was happening in their life during the term. Measured ability is one input among those. In an unselected population, general cognitive ability and academic achievement correlate substantially, which is why the relationship is real and worth stating. The evidence on that association is set out on the page on measured ability and academic achievement.

Now apply the selection. A law school class is not an unselected population. Everyone in it cleared an admissions bar, which compresses the range of the very variable you would want to correlate with rank. When the range of a predictor is restricted, its observed correlation with anything falls, sometimes sharply, without any change in the underlying relationship. The comparison inside such a group therefore answers a narrow question about who did better than whom among people who were already selected, and it does not answer a question about the distribution as a whole.

The clearest empirical demonstration of that principle in the intelligence literature is Terman's own cohort. Selected at roughly 135 and above, the men later judged most and least successful differed by five points of childhood Stanford-Binet in the 1947 volume and by 7.6 points in the 1968 reselection, while differing by a factor of four in median family income. Where everyone is far above the mean, the remaining variance in the score explains very little of the remaining variance in outcomes. The full argument, with the primary tables, is on the page on the 140 threshold.

5 The Law School Plagiarism Episode, Documented

The episode is in the file he released, he acknowledged it at the time, and it bears on academic conduct rather than on measured ability. It belongs on this page because it is part of the documented record and because leaving it out would make the record look tidier than it is.

According to the contemporaneous Christian Science Monitor account, in his first year at Syracuse he was caught taking material from a law journal in writing an assigned brief without the required attribution, and he received an F in the course. The faculty allowed him to take the course again the following year, and the second time he earned a B. The source article he drew from was a Fordham Law Review piece. Biden characterized the incident at the time as a mistake rather than an intentional deception, and the F and the retake are both in the released file.

A separate and better known episode from the same period involved passages from a speech by the British Labour leader Neil Kinnock used without attribution during the 1987 campaign, which together with the academic questions ended that campaign in September 1987. It is mentioned here only because search results collapse the two, and because the distinction matters for anyone trying to keep the documentary record straight.

Neither episode is evidence about cognitive ability, and it is worth saying why in psychometric terms rather than as a courtesy. An intelligence test estimates performance on standardized tasks under controlled conditions against an age matched reference sample. Conduct in a course, a decision under deadline pressure, and honesty in citation are not sampled by any of the tasks on such a test, and no published instrument claims otherwise. What test publishers do claim, and what is documented, is that scores predict performance on other cognitive tasks and, more weakly, on educational and occupational criteria. That boundary is what the page on what IQ tests measure exists to draw.

The same boundary is why this page carries no assessment of anyone's decisions in office. A page about measurement that starts grading behavior has stopped being a page about measurement.

6 What the Montreal Cognitive Assessment Actually Is

The MoCA is a ten minute, 30 point screening instrument built to catch mild cognitive impairment in people who look normal on a cruder screen, and it was never designed to rank healthy adults on intelligence. Because it is the only formal cognitive instrument that has appeared in modern presidential health reporting, it is the instrument people mean when they ask whether a president has been "cognitively tested."

The primary source is Nasreddine, Phillips, Bedirian, Charbonneau, Whitehead, Collin, Cummings and Chertkow, in the Journal of the American Geriatrics Society, volume 53, issue 4, pages 695 to 699, 2005. The stated objective is explicit in the abstract: to develop a ten minute cognitive screening tool to assist first line physicians in detection of mild cognitive impairment, described there as a clinical state that often progresses to dementia. The validation compared 94 patients meeting clinical criteria for mild cognitive impairment, 93 patients with mild Alzheimer's disease, and 90 healthy elderly controls.

The results are the reason the instrument spread. At a cutoff of 26 out of 30, the older Mini-Mental State Examination detected 18 percent of the mild cognitive impairment cases while the MoCA detected 90 percent. In the mild Alzheimer's group the figures were 78 percent and 100 percent. Specificity was 100 percent for the MMSE and 87 percent for the MoCA. Those numbers describe an instrument that is good at flagging impairment in an older clinical population, and they say nothing at all about discrimination among healthy adults.

The cutoff has since been revisited. Carson, Leach and Murphy screened 304 studies and meta-analyzed the nine that met inclusion criteria, reporting in the International Journal of Geriatric Psychiatry, volume 33, issue 2, pages 379 to 388, in 2018, that a cutoff of 23 out of 30 yielded better diagnostic accuracy than 26 and lowered the false positive rate, which had been inflated particularly among older and less educated test takers. A screening threshold that moves by three points on further evidence is behaving normally for a screen. It is also a reminder that the boundary is a clinical decision rule and not a measured quantity.

7 Why a Perfect Screen Score Says Nothing About IQ

A screening instrument and an intelligence test are built for opposite jobs, and the design choices that make one good make the other useless. This is the section that answers the actual question behind most searches on this topic, and the answer does not depend on who took the test.

A screen is designed to be failed only by people with a real problem. Its items are therefore easy on purpose, and almost everyone without impairment gets almost everything right. That produces a ceiling: the top of the scale is crowded, because it has to be. An intelligence test is designed to spread people out across the whole range, so its items climb in difficulty until the strongest test taker starts missing them, and the scale is anchored to an age matched reference sample so that the same number means the same rarity at every age. A screen has a cutoff. A test has a distribution.

Read the last row carefully, because it is the whole point. On a screen, a perfect score is the expected result for a healthy adult and carries almost no information beyond the absence of a flag. On an intelligence test, a perfect score is a measurement failure that tells the examiner the instrument ran out of difficulty before the test taker ran out of ability. Treating a 30 out of 30 as though it were a high IQ result reads a ceiling as a peak.

The same asymmetry applies to precision. Because a screen has few items and they are easy, one careless error moves the total by a full point out of thirty, and there is no standard error of measurement printed to put a band around it. Modern ability composites report intervals precisely because a point score overstates what was measured. Why those intervals exist and how they are derived is covered on the reliability and validity page, and the arithmetic that turns a score into a standing is on the page that converts a score to a percentile.

8 How a Dementia Screen Entered Presidential Reporting

The MoCA became a public topic because of one White House briefing, and the physician who administered it described it accurately at the time. The documentary record here is a government transcript, which makes it unusually easy to check.

On 16 January 2018, Dr. Ronny Jackson briefed reporters at the White House on the sitting president's annual physical. The official transcript records him naming the instrument, "the Montreal Cognitive Assessment, sometimes referred to as the MoCA," reporting a result of 30 out of 30, and explaining that he had not considered it clinically indicated but included it at the president's request. His own characterization of what it is could not be plainer: "It's a screening tool." He also described what it screens for, which is cognitive issues of the kind associated with dementia.

That is the entire evidentiary content of the episode, and it does not extend to intelligence. A 30 out of 30 on a dementia screen means the screen found nothing to refer. It does not place anyone on an ability distribution, because the instrument has no ability distribution to place people on. Coverage at the time frequently reported the result as though it were a general cognitive rating, which is the misreading this section exists to correct.

For Biden specifically, the published summaries from his presidential physicals did not report a MoCA result or any other cognitive screening score, and his physician stated publicly that he did not consider one clinically indicated. That is a fact about what a document contains. It is not evidence about intelligence in either direction, for the same reason a 30 out of 30 is not: an instrument that does not measure a quantity produces no information about that quantity whether it is administered or withheld. This page takes no position on the clinical judgment involved, which is outside its subject.

9 The Stutter, and Why Fluency Is Not Ability

Biden's stutter is documented, longstanding, and genuinely relevant to this page, because speech fluency and verbal ability are different constructs and a fluency disorder does not lower the underlying ability. It is also the part of the record most often misread as though it were evidence about cognition.

The record is his own. He has described the stutter since childhood, has written about it, and gave an extended interview about it to John Hendrickson for The Atlantic in the January and February 2020 issue, which remains the most detailed public account: failed speech therapy in kindergarten, and a self directed regimen as a teenager of reciting Yeats and Emerson in front of a bedroom mirror with attention to rhythm. He has also stated publicly that his stutter is not the explanation for every verbal stumble, which is a distinction worth preserving rather than flattening in either direction.

The clinical background is on the National Institute on Deafness and Other Communication Disorders page on stuttering, which describes it as a speech disorder characterized by repetition of sounds, syllables, or words, prolongation of sounds, and blocks. Roughly 3 million Americans stutter. Between 5 and 10 percent of children stutter at some point and about 75 percent recover. Boys are two to three times as likely to stutter as girls. The page also records a correction the field made about itself: stuttering was once believed to be psychogenic and caused by emotional trauma, and psychogenic stuttering is now understood to be rare.

The honest complication is that the research literature does not say "no association with anything." Ntourou, Conture and Lipsey meta-analyzed 22 studies of children between ages two and eight in the American Journal of Speech-Language Pathology, volume 20, issue 3, pages 163 to 179, in 2011, and reported that children who stutter scored lower than children who do not on norm referenced measures of overall language, with Hedges's g of -0.48, receptive vocabulary at -0.52, expressive vocabulary at -0.41, and mean length of utterance at -0.23. Those are small to moderate group level differences on early childhood language measures. They are not adult IQ scores, they are not diagnoses of anyone, and a difference of that size means the two distributions overlap almost completely. A group mean has never told anyone where an individual sits, and the authors did not claim it did.

What the psychometric structure says is more direct. On a Wechsler style verbal comprehension task, the examiner scores the content of a response against a criterion, not the smoothness of its delivery. A definition that arrives after a block scores exactly what the same definition scores when it arrives immediately. Fluency affects the channel. Ability is what is being sent through it. That distinction is what makes the next section matter.

10 What a Verbal Index Measures, and What It Does Not

Verbal ability in the psychometric sense is crystallized knowledge and reasoning with words, and it is measured by content, not by delivery. Getting this right is the difference between an argument about evidence and an argument about how somebody sounds.

In the CHC framework, the relevant broad ability is Gc, comprehension knowledge. It is estimated from tasks such as defining words, stating what two concepts have in common, retrieving general information, and identifying synonyms and antonyms. None of those tasks scores speech rate, hesitation, repetition, or blocks. A test that scored fluency would be measuring something else and would say so in its manual. The domain structure is set out on the cognitive domains page and its theoretical background on the CHC model page.

ACIS estimates Gc through the Verbal Comprehension Index, built from five subtests: Antonyms, Vocabulary, Information, Synonyms and Similarities. The published technical figures for that index are an omega reliability of .9745, a standard error of measurement of 2.40 IQ points, and a g loading of .864. At subtest level the g loadings run Similarities .865, Antonyms .860, Vocabulary .848, Synonyms .828 and Information .814. The index correlates .801 with Fluid Reasoning and .577 with Processing Speed, which is the ordinary finding that verbal knowledge and speed of simple decisions are the two least related things a battery measures. Those figures come from the published normative model in the technical manual.

Two practical consequences follow for anyone with a fluency disorder taking a verbal test. The first is that untimed verbal comprehension tasks are largely unaffected, because the score depends on the content of the answer. The second is that a timed oral task is a different situation, and this is where a fluency disorder can depress a score without any change in the underlying ability. Test standards address exactly this: an accommodation exists to keep the score a measure of the construct rather than a measure of the disorder. What a broader verbal battery covers, and where the timed elements sit, is laid out on the verbal IQ test page.

The general reason a single number hides this is that composites average across domains. A person can sit well above the mean on Gc and near the middle on Gs, and the composite describes neither accurately on its own. The index profile is what recovers the shape, which is the argument developed on the fluid versus crystallized comparison.

11 Why Estimating a President's IQ Does Not Work

Retrospective estimates of a public figure's intelligence track how much documentation survived about them, and that is not a hypothesis but a finding the method's own inventor reported. Any page offering you a figure for a living politician is running a version of this method, usually without the qualifications.

The method traces to Catharine Cox, whose 1926 volume in Terman's series estimated childhood IQ for 301 eminent historical figures from biographical records. Cox found, and reported plainly, that the more complete the surviving record, the higher the resulting estimate, a relationship she described as entirely unexpected and discovered only after the ratings had been tabulated. Her estimates correlated .25 with rank order of eminence, falling to .16 once the reliability of the underlying data was held constant, inside a group already selected for fame. The full account of that volume and its qualifications is on the page on the genius threshold.

The cross president comparison, including the reconstructions Dean Keith Simonton published, is owned by a separate page here and is not restated on this one. If you want the ranking, the method behind it, and the reasons the reconstruction cannot be applied to a recent officeholder, read the review of presidential IQ estimates. The relevant point for this page is the one Simonton himself made in a footnote: the method needs biographical material that only accumulates after a term ends, so every figure circulating for a recent president comes from a weaker source than the ones in the published series.

The same trace has been run for the two most searched contemporaries. The evidence review for Donald Trump and the evidence review for Barack Obama reach the same structural conclusion this page does, from different documents: no administration, therefore no score, and the circulating figures resolve to aggregator pages citing each other. Three reviews built independently that all end in the same place is itself informative about the genre.

There is a further reason estimates of living political figures are worse than estimates of the historical dead. The biographical record for a serving politician is generated by adversaries and supporters in real time, and every anecdote in it was selected for its usefulness in an argument. Cox at least worked from archives assembled by biographers with no stake in a current election. A method that already tracks documentation rather than ability performs worse, not better, when the documentation is produced by people trying to win something.

12 What Would Actually Count as Evidence

A defensible cognitive figure for any person requires four things, and none of them is present for any sitting or former president. Naming the four is more useful than another paragraph about what is missing.

• A named instrument with a published manual. Not "an aptitude test," not "a cognitive test," but a specific edition of a specific battery with documented norms.
• A date and an examiner. Scores drift with norm generation and with the conditions of administration, so an undated figure is uninterpretable even when it is real.
• A reference sample the score was converted against. A raw total means nothing until it is ranked against an age matched normative frame.
• A document a third party can open. A quotation from a person about their own result is a report, not a record. The distinction decided the whole Kaczynski case, where a court unsealed the administration and the widely repeated childhood figure turned out to have no sheet behind it.

Measured against that list, the Biden record contains a transcript, a class rank, a screening instrument discussion, and nothing else. That is not a gap this page can close, and pages that appear to close it have invented something.

If the underlying interest is your own profile rather than a public figure's, the distinction that matters is supervision. A proctored session with a licensed psychologist using a current battery produces a result institutions accept, because conditions were controlled, identity was verified, and a qualified examiner observed the performance. Nothing administered over the internet substitutes for that. Where those sessions are available and what they cost is set out on the page on where to take a test.

An unsupervised online assessment answers a different question and is worth taking on its own terms. ACIS runs 20 subtests across the six CHC domains for adults aged 16 to 90, in three tiers: Quick at 15 dollars with six subtests in about 45 minutes, Optimized at 30 dollars with 13, and Full Scale at 50 dollars with all 20 and the complete composite set. Access is a one time payment lasting 30 days, the free trial lasts 7 days, and there is a 5 day guarantee. It is unsupervised, it is not a clinical instrument, and it is not appropriate for diagnosis, hiring decisions, accommodation requests or society admission. The six subtest form is described on the Quick form page, and the comparison across tiers is on the adult testing page.

13 Sources Behind This Page

Every figure above is traceable, and the ones that are not traceable are labeled as fabrications rather than repeated. The full list follows so that a reader can check the page against its own sources rather than against other pages that agree with it.

• Contemporaneous reporting on the 1987 academic release. The Christian Science Monitor, 23 September 1987, source of the year by year law school class ranks, the University of Delaware standing, the scholarship correction, and the law review course episode.
• The class rank fact check. Snopes, rated true, 27 April 2024, confirming the 76 of 85 figure against reporting by the Associated Press, The New York Times and The Washington Post. Snopes blocks automated access, so the version linked here is the syndicated copy carried by Yahoo News.
• The 187 fact check. Snopes, rated false, 30 January 2024, on the claim that a 187 was announced at the 29 January 2024 White House briefing.
• The 1987 Claremont exchange. C-SPAN archive footage of the 3 April 1987 campaign stop in Claremont, New Hampshire.
• The primary MoCA validation. Nasreddine, Phillips, Bedirian, Charbonneau, Whitehead, Collin, Cummings and Chertkow, Journal of the American Geriatrics Society, 53(4), 695 to 699, 2005, doi 10.1111/j.1532-5415.2005.53221.x.
• The cutoff re-examination. Carson, Leach and Murphy, International Journal of Geriatric Psychiatry, 33(2), 379 to 388, 2018, doi 10.1002/gps.4756.
• The 2018 presidential physical briefing. Official White House transcript, 16 January 2018, National Archives copy.
• Clinical background on stuttering. National Institute on Deafness and Other Communication Disorders, National Institutes of Health.
• Language measures in children who stutter. Ntourou, Conture and Lipsey, American Journal of Speech-Language Pathology, 20(3), 163 to 179, 2011, doi 10.1044/1058-0360(2011/09-0102).
• The biographical estimation method. Catharine Cox, The Early Mental Traits of Three Hundred Geniuses, Stanford University Press, 1926, volume II of Genetic Studies of Genius.

Two things on this page could not be verified and are stated as such. The exact page count of the released Syracuse file is given as 65 pages in reporting from the period, and the file itself is not available online for direct inspection. The contents of the presidential physical summaries are described here only as to whether a cognitive screening result appears in them, because the summaries are brief documents and this page does not interpret their clinical content.

The professional framework governing all of this is explicit. The Standards for Educational and Psychological Testing (2014), published jointly by the American Educational Research Association, the American Psychological Association and the National Council on Measurement in Education, require that a score interpretation be supported by evidence for the specific use proposed, that classification decisions be reported with their measurement error, and that the limits of the normative sample be disclosed. APA standards on test use add the requirement that a person reporting a score state what it supports and what it does not, and that instruments not be used for purposes their validity evidence does not cover. A dementia screen reported as an intelligence result, and an intelligence figure produced from a biography, both fail that requirement, and they fail it in the same way.

14 Frequently Asked Questions