Ted Kaczynski's IQ: a documented 136 and an unverified 167

0 The Short Answer

The only documented intelligence test result for Theodore Kaczynski is a Wechsler Adult Intelligence Scale Revised (WAIS-R) administration recorded in a public federal court document: Verbal 138, Performance 124, Full Scale 136. Those figures appear in the forensic evaluation written by Dr. Sally C. Johnson, dated 16 January 1998 and unsealed by order of United States District Judge Garland E. Burrell, Jr. on 11 September 1998 in United States v. Theodore John Kaczynski, docket CR S-96-259 GEB.

Between 1978 and 1995 he carried out a bombing campaign that killed three people and injured twenty three others, summarized in the FBI case record as three killed and nearly two dozen injured. He was arrested on 3 April 1996, pleaded guilty in January 1998, and died in federal custody in June 2023. This page is about where two numbers came from and what each one can support. It is not about him, and nothing here is offered as an explanation of anything he did.

The complication is the 167. That figure circulates as though it were a competing measurement, and it is not. The Johnson report addresses the childhood testing directly and says the underlying records were never made available. What the report preserves is his own recollection of a range, not a verified score. Even if a 167 was written down in the early 1950s, a school score from that era and an adult WAIS-R score are computed on different scales and cannot be placed side by side. That distinction, between a ratio score and a deviation score, is the substance of this page.

1 The Document That Records the 136

Most celebrity IQ figures have no document behind them at all, which is what makes this case unusual: the primary source is a federal court record that anyone can read. The document is a forensic evaluation prepared for a competency to stand trial determination. It runs to dozens of pages, names its own sources, and was produced under adversarial conditions with retained experts on both sides.

The provenance chain matters more than the number. On 9 January 1997 the district court ordered that the defendant be examined by Bureau of Prisons physicians to determine his mental competency to stand trial. Dr. Sally C. Johnson, then chief psychiatrist and associate warden for health services at the federal facility in Butner, North Carolina, conducted that evaluation. Her report carries a date of report of 16 January 1998. A redacted copy was unsealed by court order the following September and has been publicly available since.

A competency evaluation is a narrow legal instrument. It asks whether a defendant can understand the nature and consequences of the proceedings and assist counsel in a defense. It is not an intelligence assessment, and Johnson says so plainly: she notes that the intent of neuropsychological testing is not to accomplish clinical diagnosis, and explains that for competency purposes she focused on two instruments already administered. The IQ figures are incidental to the document's purpose, which is one reason they are more trustworthy than a number produced for publicity.

Two provenance details are easy to miss and both change how the figures should be read. First, Johnson did not administer the testing. Neuropsychological testing including intellectual and memory testing was completed in May 1996 by others, and Johnson reviewed it rather than repeating it. Second, the administration date is 1 May 1996, roughly four weeks after his arrest, in custody. Testing conducted in a forensic setting under those conditions is standard practice and the report treats the results as usable, but conditions of administration are part of what a score means. The same caution applies to any unsupervised or non standard administration, which is why measurement quality is always reported alongside a score rather than after it.

2 What the WAIS-R Actually Recorded

The report gives three numbers and one interpretive sentence, and that is the entire documented psychometric record on his adult cognitive ability. The WAIS-R results were a Verbal score of 138, a Performance score of 124, and a Full Scale score of 136. Johnson adds that the split between Verbal and Performance was large but not significant, because there was, in her words, "no impairment in the performance score and no specific deficits in any subtests." She characterizes the profile as a very strong verbal ability level with a lower but still above average performance ability.

The WAIS-R is a deviation scored instrument. Published by David Wechsler in 1981, it retained the eleven subtest structure of the original scale and produced Verbal, Performance and Full Scale IQs, as described by Crawford, Allan, Stephen, Parker and Besson in their 1989 factor structure study in Personality and Individual Differences (volume 10, issue 11, pages 1209 to 1212). Wechsler IQs are placed on a scale with a mean of 100 and a standard deviation of 15, which is what makes the following conversions possible.

The percentile and rarity columns are normal curve arithmetic applied to the standard Wechsler metric. They are not printed in the report, and they should be read as approximations rather than as findings. If you want the same conversion for any other score, the exact percentile table and the one in X rarity calculator do the arithmetic without the rounding.

One further caution belongs on any point score, including this one. The report prints three integers with no confidence intervals, which is normal for a document of this kind and unhelpful for interpretation. Every observed score sits inside a band determined by the instrument's standard error of measurement. A 136 is best read as a region rather than a coordinate, and the same is true of a 138 or a 124. The report also notes a Wechsler Memory Scale Revised result that was broadly consistent with the intelligence testing, with one lower visual memory index that Johnson traced to a single subtest and did not treat as clinically meaningful for the competency question.

3 Where the 167 Comes From

The number 167 does not appear in the Johnson report, and the closest thing to it in the primary record is a self reported range that the evaluator could not verify. This is the single most important fact on the page, and it is almost never stated in the pages that rank for this query.

Here is what the report actually says. He attended fifth through eighth grade at Evergreen Park Central School. As the result of testing conducted in the fifth grade, it was determined that he could skip the sixth grade and enroll with the seventh grade class. According to various accounts, testing showed him to have a high IQ and, by his account, his parents were told he was a genius. He claims that his IQ was in the 160 to 170 range. Johnson then adds the sentence that settles the matter: "Testing supposedly conducted at that time has not been made available for review."

Parse the epistemics carefully, because they are doing all the work. The fact of fifth grade testing is documented, in that a school made an acceleration decision on the basis of it. The grade skip is documented. The high IQ is attributed to various accounts. The specific range of 160 to 170 is reported by the subject himself, decades later, in a forensic interview. The score sheets are absent. Nowhere in that chain does a specific figure of 167 appear, and a federal forensic evaluator with subpoena backed access to his records could not obtain the underlying testing.

So where did 167 come from? The honest answer is that no traceable origin has been established. It appears in press coverage, in aggregator listicles and in AI generated summaries, each of which cites the previous one. It is a stable number precisely because nothing anchors it: there is no document to check it against, so nothing ever corrects it. That pattern is the norm rather than the exception in this genre, which is why the page on record high IQ claims concludes that the famous extreme figures are generally childhood ratio scores, estimates or non standard results rather than measurements.

None of this proves he did not test high as a child. He was accelerated twice by school administrators who presumably had a reason, and the report describes strong measured verbal ability in adulthood. The claim under examination is narrower: that a specific figure of 167 is a measurement of the same thing the 136 measures. It is not, and the next two sections explain why the two numbers could not be compared even if the childhood score sheet turned up tomorrow.

4 Ratio IQ and Deviation IQ Are Different Metrics

A ratio IQ and a deviation IQ are two different calculations that happen to share a name, a scale midpoint of 100 and nothing else. Putting a childhood ratio score beside an adult deviation score is not a close comparison. It is a category error, in the way that comparing a temperature in Fahrenheit to one in Celsius is a category error even though both are numbers with degrees attached.

The ratio formula comes from Lewis Terman's 1916 The Measurement of Intelligence, the book that introduced the Stanford revision of the Binet Simon scale to American schools. Terman's own definition is compact: the intelligence quotient is "the ratio of mental age to chronological age." In practice, a child's performance is converted to a mental age by finding the age level whose tasks they can pass, that mental age is divided by their actual age, and the quotient is multiplied by 100. A ten year old credited with a mental age of twelve receives a 120.

A deviation IQ does something structurally different. It ignores mental age entirely, ranks the person against a reference sample of people their own age, and expresses the result as a distance from that group's average in standard deviation units. The Wechsler scales fix that standard deviation at 15, so a 136 means the same distance from the age matched mean whether the person is seventeen or seventy. Modern reporting works this way for every major instrument, which is what makes age based norming the load bearing part of any published score. Age based norming is also why a childhood standing does not fix an adult one: the Lothian cohorts, retested on the same test they sat at eleven, produce a rank order stability coefficient of .67 from age 11 to age 70 and .51 from 11 to 87, which preserves most of an ordering while leaving a great deal of individual movement inside it.

The two rows that matter most for this page are the last three. A ratio score has no fixed relationship to rarity, it misbehaves at the top of the range, and it cannot be computed for an adult at all without an arbitrary patch. That is not a modern criticism applied retrospectively. Terman built the patch himself, and the next section shows what it was. Our page on the history of IQ scoring follows the same transition across the rest of the field.

5 Why Ratio Scores Inflate for Young Children

The ratio formula divides by the child's age, so the younger the child, the more a fixed amount of advancement is worth. This is arithmetic, not a subtle psychometric artifact, and it can be demonstrated in three lines.

Take a fifth grader of ten years and six months who is credited with a mental age of seventeen years and six months. Divide 17.5 by 10.5 and multiply by 100 and you get 167. Now hold the performance completely constant and change nothing but the birthday.

Identical performance, three different scores, spanning fifty eight points. A deviation IQ cannot behave this way, because it never divides by age in the first place. This is the whole reason the field abandoned ratio scoring, and it is why a childhood figure from the 1950s carries no information about where an adult sits in an adult distribution.

There is a second problem inside the first. A mental age of seventeen and a half exceeds the ceiling Terman himself imposed. Writing in 1916 about how to compute the quotient for adults, he observed that measured intelligence appears to improve little after the age of fifteen or sixteen, and concluded that any person over sixteen years of age, however old, should for purposes of calculating the quotient be considered to be just sixteen. The scale's designer capped the denominator at sixteen because the formula produced nonsense past that point. A ratio score in the 160s asks the numerator to go somewhere the scale was never built to measure.

A third issue compounds both. Under ratio scoring the spread of scores was not constant across ages, so the same printed number marked off a different slice of the population depending on how old the child was. Deviation scoring fixed this by definition. That is why a bare number with no scale attached cannot be converted to a rarity: on a scale with a standard deviation of 15, a 167 sits about 4.5 standard deviations out, roughly one in a quarter of a million by normal curve arithmetic, while on a scale with a standard deviation of 16 the same 167 is about one in seventy thousand. Both figures are illustrative rather than claims about any person, and both exceed the range in which any standard adult instrument is normed. Our page on the gifted score range covers what happens to precision as you move into that territory. The same missing dispersion is recorded in the ratio figures Simonton inherited from Cox for forty two presidents, where a footnote states that those scores carry no preset standard deviation of 15 or 16 and that the sample value tends to run between 14 and 15.

6 Which Test the School Used Is Not in the Record

The most defensible statement about the childhood score is that nobody knows what instrument produced it, and the metric therefore cannot be identified. Pages that confidently label the 167 a ratio IQ are probably right about the era and are still asserting more than the evidence supports.

Here is the difficulty. The testing took place around 1952 or 1953, given his birth date of 22 May 1942 and fifth grade placement. Two families of instrument were in school use at that time and they scored differently from one another. The Stanford-Binet lineage descending from Terman's 1916 revision reported a ratio quotient. The Wechsler Intelligence Scale for Children, published in 1949, reported deviation scores from the outset, following the approach Wechsler had already taken with adults. A fifth grade screening in a suburban Illinois district in the early 1950s could plausibly have been either.

The Johnson report does not name the instrument, the examiner, the date or the score. It says the testing was not made available for review. In the absence of that record, the instrument is unknown, and with it the metric, the norm sample, the standard deviation of the scale and therefore the meaning of any number attached to it.

This is not a technicality that can be waved through. The four facts you need in order to interpret any score are the instrument, the norms it was scored against, the date of administration and the conditions. A number stripped of all four is not a weak measurement. It is not a measurement, in the sense that nothing can be inferred from it, and the honest label is unsupported. The same reasoning is why our reviews of free versus validated instruments treat a score without a documented norm table as uninterpretable rather than merely imprecise.

What can be said is narrower and still worth saying. A school district made two acceleration decisions about him, skipping him past sixth grade and later past eleventh grade, and school districts do not usually accelerate children twice without test evidence. The acceleration is documented in the report and is real. It supports a conclusion about what his school believed in the early 1950s. It does not license a number, and it certainly does not license a number that can be compared with an adult WAIS-R result recorded forty years later. Acceleration is an administrative decision, and treating it as a proxy measurement is the same error as reading a score off a biography, which is something this site does not do.

7 The 1996 Testing Used Norms That Were Already Old

Even the documented figure carries a qualification that almost no coverage mentions: the WAIS-R was administered in 1996 against a norm sample collected between 1976 and 1980. Scores obtained on aging norms drift upward, which means the 136 is, if anything, mildly generous relative to the population of the year it was administered.

The norming window is documented. Crawford and colleagues, in the 1989 Personality and Individual Differences study cited earlier, record that the WAIS-R was standardized on a representative sample of 1,880 Americans during 1976 to 1980 and published by Wechsler in 1981. The Johnson report records the administration date as 1 May 1996. The gap between the norming midpoint and the testing is roughly eighteen years.

The drift rate is also documented. Trahan, Stuebing, Fletcher and Hiscock published a meta-analysis of the Flynn effect in Psychological Bulletin in 2014 (volume 140, issue 5, pages 1332 to 1360), covering 285 studies with a combined N of 14,031. Their overall estimate was 2.31 standard score points per decade, with a 95 percent confidence interval of 1.99 to 2.64. For the subset of 53 comparisons using modern Stanford-Binet and Wechsler instruments, the estimate was 2.93 IQ points per decade, confidence interval 2.3 to 3.5.

Applying the modern instrument rate across eighteen years gives roughly five points of expected drift, with a plausible band of about four to six. On norms current in 1996, a Full Scale 136 obtained against 1978 vintage norms would correspond to something in the low 130s. State the limits of that inference plainly: this is arithmetic applied to a published population drift rate, not a rescoring. Nobody has rescored his 1996 protocol against later norms, no corrected figure exists, and it would be wrong to present one as though it did. The point is directional, not precise, and the direction is that the documented number is not an underestimate.

This is a general problem, not a quirk of this case. Any score you encounter is a score against a particular norm generation, which is why secular score gains matter whenever old and new figures are compared, and why the current edition of the Wechsler adult scale exists at all.

8 Reading the Verbal and Performance Split

The fourteen point gap between the Verbal 138 and the Performance 124 is the most interesting thing in the record, and the report is careful about what it means. Johnson describes the split as large but not significant, on the grounds that the performance score showed no impairment and no specific subtest deficits, and reads the profile as strong verbal ability with lower but still above average performance ability.

That reading is conservative and correct in its own terms. A split becomes clinically interesting when one side falls into a range that suggests a problem, or when the pattern within subtests is uneven. Neither applied here. A 124 sits around the 94th percentile. There is nothing to explain.

The general lesson is worth more than the specific case, because the intuition that a gap must mean something is very strong and usually wrong. Index scores are themselves aggregates, and aggregates differ from one another for uninteresting reasons: measurement error in both directions, subtest content that suits one person's background, differences in test taking speed. Two indices four or five points apart tell you nothing at all. Two indices twenty five points apart, with a consistent pattern underneath, are worth reading.

Modern instruments have moved away from the Verbal and Performance dichotomy the WAIS-R used. The old Performance scale mixed several distinct abilities that are now reported separately, so mapping a 1981 vintage Performance IQ onto any single modern index is not sound. Contemporary batteries structured on the Cattell Horn Carroll framework separate Gc comprehension knowledge from Gv visual spatial ability, Gf fluid reasoning, Gwm working memory, Gq quantitative ability and Gs processing speed, and report an index for each. A high verbal score and a merely good visual spatial score are two different findings under that structure, where the WAIS-R would compress both into a two number summary.

ACIS reports on that six domain structure: 20 subtests across six primary cognitive domains, with subtest scaled scores on a mean of 10 and standard deviation of 3, and composites on a mean of 100 and standard deviation of 15. The Full Scale composite has an omega of .9886 and a g loading of .958, with a standard error of measurement of about 1.60 IQ points, computed on a technical analysis set of 2,750 complete records. Those figures come with real limits: the sample is self selected rather than census based, administration is unsupervised, and the adult reference frame of 3,243 English speaking records aged 16 to 90 is a modeled frame rather than a stratified national sample. The full derivation is in the technical manual.

9 Harvard at Sixteen and the Michigan Doctorate

The educational record in the Johnson report is well documented and frequently garbled elsewhere, so it is worth setting out exactly as the document has it. None of it is a measurement, and none of it should be read as one.

He attended kindergarten through fourth grade at Sherman Elementary School in Chicago, then fifth through eighth grade at Evergreen Park Central School, where the fifth grade testing led to the sixth grade being skipped. He attended Evergreen Park Community High School. The report says he did well academically overall but had some difficulty with mathematics in his sophomore year, was subsequently placed in a more advanced mathematics class, mastered the material, and then skipped eleventh grade. The two skips together meant he completed high school two years early, though this required a summer school course in English.

He was encouraged to apply to Harvard in his later high school years, was accepted, and began in the fall of 1958 at the age of sixteen. He completed an undergraduate degree in mathematics, graduating in June 1962 at the age of twenty. He began graduate study at the University of Michigan at Ann Arbor in the fall of 1962 and completed a master's degree and a doctorate in mathematics by the age of twenty five. Following graduation he took a position as assistant professor in the mathematics department at the University of California at Berkeley, holding it from September 1967 until June 1969.

Notice what this record does and does not contain. It contains dates, institutions, degrees and ages, all verifiable. It contains one detail that cuts against the popular telling, which is that he struggled with sophomore mathematics before being moved up. It contains no test scores whatsoever. Educational attainment and measured ability are correlated at the group level and the correlation is far from perfect, which is the subject of our page on IQ and academic achievement. Running that correlation backwards to infer an individual's score from their degrees is exactly the inference this site refuses to make, for any subject, however tempting the biography.

Grade acceleration in particular is a weak signal. It reflects a district's policy, a family's willingness, a teacher's judgment and local resources at least as much as it reflects a test result, and the report itself records that he experienced the sixth grade skip as socially costly and described it as a pivotal event in his life. Whatever else acceleration is, it is not a score, and a reader who wants to know what a score would have shown has only the 1996 WAIS-R to work from.

10 The Murray Study and the Surviving Test Data

As a Harvard undergraduate he took part in a psychological research study, and the surviving paperwork from it is the earliest psychometric material the 1998 evaluation could actually obtain. This section reports what the record says about the study and stops there. It does not speculate about effects, and the causal claims that circulate about this episode are not established.

The study is named in the Johnson report's list of source materials as the Multiform Assessment of College Men Study, conducted by Henry A. Murray at Harvard University. The report notes that he wrote an autobiography in connection with participating in it, that the participation began in his sophomore year, and that the study examined the psychological functioning of young men at Harvard. Murray was a prominent Harvard personality psychologist. The Johnson report treats the study purely as a source of archived records, and this page does the same.

What the 1998 evaluation could recover from it was limited and specific. Johnson writes that limited testing was available from Harvard where he had been involved in the Murray Study, and that the opportunity existed to review a Minnesota Multiphasic Personality Inventory profile. She observed that the Si scale, measuring social introversion, had not been scored at the time. Because a copy of the original answer sheet had been preserved, she was able to score it, and found a marked elevation on the introversion scale with a lesser elevation on the depression scale. The report also refers to projective testing from that period using the Thematic Apperception Test.

Three observations follow, and all three are about evidence rather than about him. The earliest surviving psychometric record on him is a personality inventory, not an intelligence test, which is a further reason the childhood IQ figure has nothing behind it. An unscored scale sat in a file for nearly forty years until someone thought to score it, which is a useful reminder that archived test data is only as good as the analysis someone eventually performs on it. And a 1959 MMPI profile is scored against 1940s norms, carrying the same norm generation problem discussed above.

11 What a Test Score Does Not Explain

A cognitive test score describes performance on a defined set of tasks under defined conditions at one sitting, and that is the entire claim it makes. It does not explain conduct, predict conduct, or bear on culpability, and no reading of the record supports treating it as though it did.

The document these figures come from was written to answer one narrow legal question. Johnson concluded that he was able to understand the nature and consequences of the proceedings against him and to assist his attorneys, and therefore competent to stand trial. That is a finding about capacity to participate in a legal process on a particular date. It is not a finding about why anything happened, and the report does not present it as one.

The same report also records diagnostic impressions under the DSM-IV framework then in use, listing a provisional Axis I diagnosis of paranoid type schizophrenia, episodic with interepisode residual symptoms, and an Axis II impression of paranoid personality disorder with avoidant and antisocial features. That diagnostic material was contested at the time by experts retained by the prosecution, who reached different conclusions from a review of materials without an interview. This page reports the existence of that disagreement and adjudicates none of it. A diagnosis is not an explanation either, and the presence of both a diagnosis and a high test score in one document does not create a relationship between them.

There is a broader failure mode worth naming, because it is what drives traffic to pages like this one. Readers encounter a high score attached to a person who did terrible things and feel that the combination demands an account. It does not. Cognitive ability scores show modest group level associations with educational and occupational outcomes and effectively no useful individual level relationship with conduct. Our page on IQ and crime research works through the cohort evidence, the confounding, and the effect sizes involved. Its conclusion is the one that applies here: those are population averages with substantial confounds, and a group statistic can never be run backwards onto an individual case as a diagnosis or an explanation.

12 How to Check Any Reported IQ Figure

The method that separates the 136 from the 167 is general, and once you have run it a few times most celebrity IQ figures collapse in under a minute. Ask four questions, in order, and stop as soon as one of them has no answer.

First, what instrument produced the number. Second, what norm sample was it scored against and when was that sample collected. Third, when and under what conditions was it administered. Fourth, can you open the document that records it. A figure that survives all four is a measurement. A figure that fails any of them is a claim, and should be labeled as one.

Two failure patterns account for most of what circulates. The first is metric confusion, where a ratio score, a high range test result or a score from an instrument with a different standard deviation is quoted as though it were a modern deviation IQ. The second is citation laundering, where an unsourced number is repeated by a second outlet that credits the first, and by a third that credits the second, until the repetition itself starts to look like corroboration. Neither pattern requires anyone to lie, which is why both are so durable, and why the common myths about IQ testing tend to survive correction. Across a larger set of names both patterns are the norm rather than the exception: the audit that sorts famous IQ claims into four evidence classes places twenty five of its thirty five cases in the class with no identifiable origin, and four in the class with any administration behind them at all.

Apply the four questions to your own score too, whether it came from a clinician, from a free web quiz or from us. If a test cannot tell you its norm sample, its reliability and its standard error, then it has told you a number and not a measurement. That is the standard we hold ourselves to on the accuracy page and it is the same standard applied to the two figures above.

13 Profiles, Provenance and Testing Standards

The reason this case is instructive is that it contains, in one document, almost every provenance problem that makes public IQ figures unreliable, plus one properly recorded result to compare them against. A single number without its instrument, norms, date and source is not a small measurement. It is not a measurement.

It also shows why a profile carries more information than a composite. The most informative thing in the record is not the 136 but the shape underneath it, a Verbal 138 against a Performance 124, which the evaluator read as strong verbal ability alongside solid but lower non verbal performance. That shape survives when the composite is uninformative. Contemporary reporting goes further, separating what the old Performance scale merged, and giving Gf fluid reasoning, Gc comprehension knowledge, Gq quantitative ability, Gv visual spatial ability, Gwm working memory and Gs processing speed their own indices. Our page on what scores mean in practice covers how to read those indices against each other rather than against a single headline figure.

ACIS is built on that structure: 20 subtests across six primary cognitive domains, producing six index scores alongside a Full Scale composite. The composite's psychometric properties are published rather than asserted, with a g loading of .958 and a higher order confirmatory model fitting at CFI .9761, TLI .9726, RMSEA .0406 and SRMR .0217, chi-square 916.703 on 166 degrees of freedom, estimated on 2,750 complete records. Those are real figures with real boundaries. The normative sample is self selected rather than census based, administration is unsupervised, and the adult reference frame of 3,243 English speaking records aged 16 to 90 is a modeled frame. ACIS is not a clinical instrument and is not appropriate for diagnosis, hiring decisions, educational accommodations or high IQ society admission, which is set out in full on the supervised versus online comparison. Researchers and educators who need verified scores with participant links and CSV export work through ACIS Professional instead.

All of which lands on the same professional principle. The Standards for Educational and Psychological Testing (2014), issued jointly by the American Educational Research Association, the American Psychological Association and the National Council on Measurement in Education, put the burden of evidence on the party making a score claim, and require that a score be reported with the evidence supporting the specific interpretation being proposed for the specific use at hand. The APA maintains its summary of those testing standards, with open access front matter available from the Standards project site. Under that framework the 136 is a documented score with a known instrument, a known norm sample and a readable source, carrying the qualifications this page has attached to it, and the 167 is an unsupported claim. Applying that rule consistently, to famous strangers and to your own results alike, is the whole discipline.

14 Frequently Asked Questions