In his book, A Question of Intelligence: The IQ Debate in America, the late journalist Daniel Seligman describes a four hour conversation he had with the late Arthur Jensen.

Seligman couldn’t resist asking the man who had not only launched the IQ debate in America, but dominated it for decades, what his own IQ was. From page 63 of the book:

I once asked Jensen if he knew his own IQ. It turned out that he had never taken any of the standard tests, like the WAIS. The question of testing him first arose during the year of his Maryland internship, but by then he could not take the WAIS because he was too familiar with it (having administered it to others perhaps a hundred times). Of the various mental tests he has taken over the years, the Terman Concept Mastery Test (CMT) __ a high-level measure of verbal skills__probably provides the best approximation of an IQ test. Jensen took it when he was forty-three. He declined to tell me the score__and seemed distinctly unhappy at my interest in the subject__but did finally mention that his CMT score was about at the average of those members of Terman’s Gifted Group who had gone on to earn Ph.D.s.
Poking my nose into volume 5 of Terman’s Genetic Studies of Genius, I learn that this subgroup of the gifted had Stanford-Binet IQ equivalents of 156, well into the 99.9 percentile. Which possibly helps to explain why Jensen has been such a dominant figure in the IQ debate.

Because 156 is a truly stratospheric IQ, even for a man as intellectually influential as Jensen, I wanted to know more. And how would a test given to adults be assigned a Stanford-Binet equivalent, when the Stanford-Binet (at the time of the Terman study) was scored using age ratio scores (a method that only works up to adolescence).

Luckily, I found a (poorly formatted) copy of volume 5 of Terman’s Genetic Studies of Genius. Table 15 shows the adult CMT scores of the Terman Gifted Group who obtained PhDs. Contra Seligman, the average score for this group appears to be 159, not 156, and it appears to be a raw score, not an IQ equivalent.

What IQ does a raw score of 159 on the CMT equate to? Ideally, you’d want a sample of the general U.S. (white) population to take the CMT, but since general population samples are hard to get, and are often too small to norm rare scores, psychometricians (including Ron Hoeflin) often use a technique called score-pairing, where a sample of convenience takes two tests, one of which has already been normed, and it’s assumed that the IQ distribution of the test with known IQ equivalents will match the IQ distribution of the test with unknown IQ equivalents. This assumes of course that relative to the general population, the convenience sample (as a group, not necessarily as individuals) would score equally high, and equally variable on both tests. This assumption is justified in cases where there’s no reason to assume a particular convenience sample would have an advantage on one test over another.

In a 1985 study, 150 Berkeley students were tested on both the CMT and the Advanced Progressive Matrices (APM). Their raw scores (means and standard deviations) on the CMT and APM (set II, time limit up to an hour and 15 minutes) were M = 81.69, SD = 32.8 and M = 27.24, SD = 5.14, respectively. From these equivalencies, score pairing suggests that Jensen’s CMT score of 159 (+2.36 SD relative to the Berkeley bell curve) equates to an APM score also +2.36 SD relative to the Berkeley bell curve:

The APM equivalent of Jensen’s CMT score = 2.36(Berkeley APM SD) + (Berkeley APM Mean)

The APM equivalent of Jensen’s CMT score = 2.36(5.14) + 27.24

The APM equivalent of Jensen’s CMT score = 39

In other words, Jensen’s CMT score equates to an APM score that is 3 raw score points above the APM’s ceiling of 36!!! This does not mean Jensen would have hit the ceiling on the APM of course (most high scores on one test regress a lot to the mean on the other); rather it means that if he scored as well on the APM as he did on the CTM, he’d be above the APM ceiling (under a 75 minute time limit).

Thus, to determine Jensen’s IQ equivalent on the CTM, we must determine the IQ equivalent of an APM score of 39.

In the same study, another group of Berkeley students took both the APM and the WAIS. Their scores were M = 28.23, SD = 5.08, and M= 122.84, SD = 9.3. Because the WAIS scores were expressed as IQs, not raw scores, we can use score pairing to assign WAIS IQ equivalents to APM scores, but before doing so, we should note that WAIS norms were several decades old by 1985 (in the Minnesota transracial adoption study, parents in 1985 scored 5 points lower on an abbreviated WAIS-R than they had on an abbreviated WAIS ten years earlier, suggesting the WAIS was 5 points to generous by 1985). So deducting 5 points from their WAIS IQs of Berkeley students, score pairing tells us that APM set II score (75 minute time limit) of 28.23 = IQ 117.84, and every 5.08 points above or below 28.23, add or subtract 9.3 IQ points.

From here, we can estimate that a 39 on the 75 minute APM in 1985 (and thus a 159 on the CMT) equaled a WAIS IQ of 138.

One problem though is Jensen took the CTM at age 43 circa 1966 while the Berkeley students took in their early 20s (on average) circa 1985. On the one hand this might give Jensen an unfair advantage since middle aged adults do slightly better on verbal tests than younger adults. On the other hand it might have given him a slight disadvantage since older birth cohorts sometimes do slightly worse on verbal tests than newer cohorts. Likely both factors cancel out.

A king among men

To realize just how impressive an IQ of 138 is, it is useful to compare it to our best and brightest: the students at the most elite universities in the entire World:

After correcting for old norms, a sample of Harvard students averaged IQ 122 on an abbreviated version of the WAIS-R . After correcting for old norms, a sample of Dartmouth students averaged IQ 122 on the WAIS. Meanwhile a sample of Oxford students apparently averaged IQ 116 on the Advanced Progressive Matrices (APM). On average, studies of the most elite university students show a mean IQ of 120.

Of course It’s likely this is an underestimate. The samples who volunteer for such studies might be disproportionately drawn from psychology students, leaving higher IQ STEM students underrepresented. Some tests may not have sufficiently high ceilings in all domains for the brightest students in the samples to show their full ability. Tests like the APM might be biased against more verbally oriented intellects. In addition, correcting for old-norms may over-correct, because the Flynn effect has been exaggerated in my opinion.

Given all this, perhaps the true average IQ of the World’s most elite universities is actually 125-130. Even still, with an IQ of 138, Jensen could have walked into any one of them, and cognitively towered over most of the students and professors.

He was not just a king among men, as the journal Intelligence christened him, but a king among the intelligentsia. A man who took a field as fluffy as psychology, and turned into rigorous science. There’s no one who even comes close; no heir apparent on the horizon.

“He’s probably a Genius,” gushed the late J.P. Rushton, who many regard as a Genius himself.

Even Geniuses make mistakes

And yet for all Jensen’s Genius, he did make three mistakes in my humble opinion, reminding us that even IQ 138 minds have their limits. The first and most egregious was citing regression to the different population means (when predicting offspring IQ from the IQ of parents) as evidence of racially genetic IQ differences (see section nine of a paper he co-wrote). Regression to the mean is caused by imperfect correlations, it has nothing to do with genetics per se. Charles Murray understood this in his book Coming Apart, but so many HBDers do not.

The second mistake (as described by commenter Mug of Pee) was failing to adequately consider reaction norms in any of his writings, even though it could potentially alter the interpretation of heritability studies. Time will tell if this mistake was a serious one.

Lastly, when he was first told about the Flynn effect, he dismissed it as artifact of culturally loaded tests and predicted it would be less apparent on culture reduced tests like the Raven Progressive Matrices. It turned out the Raven showed the greatest Flynn effect of all, according to James Flynn, so Jensen was not just wrong, but the opposite of right. Of course I suspect the Flynn effect on the Raven has been exaggerated.

If Jensen’s is the only book you ever read, that will be enough

When people visit my home, the first thing they do is going running through the halls like a kid in a candy store searching for the library. “My God, you know so much,” one disappointed woman told me, “I was expecting to find the library of Congress in your basement”

Soon she was sobbing: “I just want to know what you know, I want to see what you see, I want to be where you’ve been mentally.”

The truth is I haven’t read a lot of books, I explained. I’ve read one book a lot. And I hand her an old crumbling copy of The g Factor, by Arthur Jensen.

Advertisements