« Born to Run | Gene Expression Front Page | Art vs. Food??? »
April 16, 2003

Ultra-high IQ societies

Ultra high IQ societies like the Mega Society are an interesting development - they were formed by Mensa members who thought that Mensa was basically a f*** club for the high IQ and not sufficiently intellectual. Many prominent members of these societies are promoting the idea of Power tests, that is, IQ tests with no time limit where the score you get depends on how many questions you can get through. The questions become progressively more difficult so that average intellects can't get through them no matter how long they spend. I've noticed that some Mensa puzzle books have started using these tests as well. But how much scientific basis is there for them? Any readers qualified in psychometrics have any idea?

Some of the prominent ultra high IQ people seem to be quite prolific designers of these tests (indeed you might say some, like Ron Hoeflin make their living solely out of this) but have attained their own high IQ scores on normal speed tests, and others have passed both with flying colours. At least at the extreme high end people who do extremely well at power tests do extremely well at normal speed based tests too - which I suppose is the point since power tests were originally designed to accomodate the measurement of ultra high iq. Here is a discussion of the merits of power tests vs speed tests by a prominent member of this community.

Another curious thing about ultra high IQ societies is that, to put it bluntly, they seem to attract lots of people, who are, er, otherwise undistinguished in other respects, sometimes even dysfunctional[1]. Is this, as logic suggests, because such people self-select themselves to join such groups (whereas other ultra high IQ people who become distinguished intellectuals and mix with people of similar IQ don't need to join such groups) or is society really alienating and wasting a large stock of human capital because its educational system and social customs make them into misfits?

[1] Chris Langan, billed on American TV news as the 'smartest man in American' (IQ 195) and public face of the Mega Society, works as a bouncer and promotes his obscure and crackpot-sounding CTMU theory which might as well be recycled Leibniz. Marilyn Savant, billed in the Guiness Book of Records for highest IQ in the world (228) writes puzzle columns for the Sunday papers! What a waste ...

Posted by jason_s at 09:24 AM




mensa was not a fuck club when i was in. most of the people were plain looking, middle aged folks. and they actually did spend most of their time doing seriously geeky math and language problems.

i have to disagree that the smartest people tend to feel the need to associate primarily with other very smart people, whether from an IQ society or from work. it's more like they prefer to be friends with folks they get along with.

"IQ tests are not going to be used much longer."

nah. they're used in public schools, in the army and navy and air force, even in the NFL.

"or *the* brain property, g"

the brain does a lot more than the functions measured by g.

Posted by: jody at April 16, 2003 10:03 AM


I like Mensa functioning as a singles club... I joined it to meet men, and it worked. I met my current boyfriend at a Mensa TGIF.

(I do love the concept of a Mensa TGIF drinking get together... we have the brain cells to spare!)

Posted by: Jacqueline at April 16, 2003 02:03 PM


most IQ tests lose measurement value above an IQ of X.

now, replace X with your own IQ :)

Posted by: razib at April 16, 2003 03:49 PM


"Those people generally happen to be of high IQ as well"

mostly they just have similar interests, and personalities that don't clash. the difference in IQ can be considerable.

"assortative mating"

i think this effect is weaker than you hold it to be. also, as IQ rises high, it's effect is even weaker.

"IQ tests *themselves* aren't used in public schools or for employment any more."

i never even heard of an employer giving IQ tests, but i'm certain public schools and the military use them. my school district uses them in elementary school to identify the students who score above 140. the army and navy use them to determine who can do what. you can't be a green beret or SEAL without a certain score on an IQ test.

"but g is a quantitative variable that can be predicted by direct measurements."

seems to me other brain functions can also be measured directly.

paul's method only works on twins. although at some point it might be possible to estimate the potential abilities of any brain just by looking at it. but due to plasticity, looking just at the genes a brain came from will always be a less accurate way to predict. still important, but rougher, no matter how advanced our understanding of genetics.

"It's also probably the single most important characteristic of the brain."

that's debatable.

Posted by: jody at April 16, 2003 06:05 PM


Leibniz was pretty much completely discredited as early as the 18th Century (see Voltaire's Candide).

Posted by: nietzsche at April 16, 2003 06:45 PM


Leibniz is hardly discredited, disparaged yes, discredited no.

Posted by: martin at April 16, 2003 10:55 PM


i think this effect is weaker than you hold it to be. also, as IQ rises high, it's effect is even weaker.

The correlation between the IQ of two parents is 45%. That's some substantial assortive mating going on.

mostly they just have similar interests, and personalities that don't clash. the difference in IQ can be considerable.

The correlation I have seen for best-friends is, I think, 54%.

Godless: "It's also probably the single most important characteristic of the brain."

Jody: "that's debatable."

From what we know, IQ (g) is the single most important characteristic of the brain. That is b/c it can be reliably measured, has substantial physiological correlates, and predicts life outcomes (occupational/educational/basic life tasks) better than any other single measurable characteristic. Godless hasn't made any dogmatic statements- we're open to the future -but for now g plays what looks to be the most important role in human intelligence. That is in no way to say there aren't other very important actors, or that g will necessarily always be as relevant as it is now.

Posted by: Jason Malloy at April 17, 2003 07:47 AM


Godless, if you don't do a data flood here, could you please point me to where I can find the data on IQ and how it relates to things like occupation, income, etc.? Thanks.

Posted by: Jacqueline at April 17, 2003 10:16 AM


Many years ago, a friend of mine described MENSA as the "society for overintelligent underacheivers".

I've never joined, though I suspect that in some places MENSA functions as a singles bar, even places where they're mostly "plain-looking, middle-aged folks". After all, these days, there are lots of divorced middle-aged folks looking for partners. Maybe they were smart enough to hide that from Jody.

I think that people who are really intelligent and also very successful stay away from groups like MENSA for several reasons. 1) Lack of time. All that success takes work. 2) Lack of common values or interests. People who are very intelligent and not wildly successful tend to not share the same values as people who are very successful; the very successful generally value social skills much more highly than the smart but moderately successful. (They should - that's how they got rich in the first place.) 3) I don;t know if this is true of MENSA, but I've seen a lot of resentment of the very successful by the very smart. If I were very successful, I wouldn't go to a group that might be full of people who would resent me for that success, unless someone told me that *this* group wasn't that way.

Jason claims that IQ "predicts life outcomes (occupational/educational/basic life tasks) better than any other single measurable characteristic."

Perhaps better than any other single *measurable* characteristic, but I think there is an observable characteristic which predicts monetary and mating success *better* than pure intelligence: Social skills. In particular, the skills related to sales and marketing. In anything resembling a capitalist society, the ability to convince others to give you resources in exchange for what you have to offer is *more* valuable than great intelligence applied to other skills.

Digressing a little - this is probably what accounts for the great deal of resentment that many very smart people feel towards "businessmen, salesmen, marketing people" and capitalism in general. The world is run by people who are +1 sigma to +2 sigma in intelligence, while those who are +2 sigma or higher work for them. That's because many very smart children don't develop the social skills necessary to be really successful. Their teachers certainly don't encourage it, and quite often their parents don't, either. Teachers reward intelligence, not social skills. Kids who are greatly rewarded in school who don't acquire those social skills come in for a hard landing once they're out in the real world.

Posted by: Anthony at April 17, 2003 02:08 PM


I'm reworking my comments above into an entry on my blog.

Another possible cause for the lack of correlation between IQ and financial success at the top levels is that many very smart people have various mental illnesses (especially depression) which prevent them from functioning as highly as they could.

Perhaps the brain is like a computer, that when "overclocked" is prone to subtle breakdowns and errors.

Posted by: Anthony at April 17, 2003 02:26 PM


"The correlation between the IQ of two parents is 45%."

what's that mean? could you give an example?

"The correlation I have seen for best-friends is, I think, 54%."

and that means?

"From what we know, IQ (g) is the single most important characteristic of the brain."

it's important. it's not the end all, be all though. it's also a measured as a combination of different brain functions, so the brain does not really do something called g.

"That is b/c it can be reliably measured,"

as can other brain functions

"has substantial physiological correlates,"

as do other brain functions

"and predicts life outcomes (occupational/educational/basic life tasks) better than any other single measurable characteristic."

it predicts education pretty good, but is rougher for jobs, and is not close for basic life tasks.

also, in real life, we're not limited to just one measureable characteristic when predicting people's future.

"Don't make me data flood you again like I did with Jewish scientific accomplishments"

basically you flooded me with jewish supremacist bullshit, so i stopped responding.

Posted by: jody at April 17, 2003 03:01 PM


"most IQ tests lose measurement value above an IQ of X.
now, replace X with your own IQ :)"

Very clever Razib, but you obscure an important point. The correlation of anything with g falls as g increases. An IQ below 50 is a pretty sure sign of retardation or of the desire to appear retarded. An IQ above 110 makes high-school graduation very likely and prison or teen pregnancy unlikely. An IQ of 160 predicts nothing significant that an IQ of 140 does not. I say this as some-one with an IQ well over 160 according to some tests. It's important to note that many tests only go to 130 or to 145 or 155 for a reason, and that those which go higher have have weak correlations with one-another at their high ends.
Also, the claim that g correlates positively with success at all tasks is inconsistant with basic evolutionary theory (unless it carries some non-behavioral physiological cost, but studies show that it doesn't carry an obvious one). In our ancestral environment the total selection against higher than average IQs must have equalled that against lower than average IQ or the average would have risen until that was the case.
Obviously the alleles associated with large brain size had costs which don't apply today, but the other alleles that correlate with IQ must also have had costs. Any suggestions what they might have been?

Posted by: michael vassar at April 17, 2003 05:34 PM


"Basically it measures the slope of the predictor line when predicting the random variable Y by a measurement of X."

Godless, you *know* better than that. Correlation Coefficient measures the fit of the data to a predictor line, not the slope of the best-fit line.

Which means the statement "correlation between IQ of two parents" is missing some data - is the slope of the best-fit line 1.0? Is the y-intercept of the best-fit line zero? One assumes they are, but that's not obvious.

Posted by: Anthony at April 18, 2003 09:19 AM


"Also, the claim that g correlates positively with success at all tasks is inconsistant with basic evolutionary theory (unless it carries some non-behavioral physiological cost, but studies show that it doesn't carry an obvious one). In our ancestral environment the total selection against higher than average IQs must have equalled that against lower than average IQ or the average would have risen until that was the case. "

In wealthy societies, the "physiological cost" of high intelligence is fewer children. There seems to be some correlation between high intelligence and symptoms of mental illness, which can lead to low reproroductive success.

Posted by: Anthony at April 18, 2003 09:24 AM


Anthony writes
'Also, the claim that g correlates positively with success at all tasks is inconsistant with basic evolutionary theory (unless it carries some non-behavioral physiological cost, but studies show that it doesn't carry an obvious one). In our ancestral environment the total selection against higher than average IQs must have equalled that against lower than average IQ or the average would have risen until that was the case. '

Some things to consider:
Maybe the average has been rising (until recent times when successful people started having fewer children).
Maybe highly intelligent people burn more calories thinking. I have seen a study that suggested greater glucose metabolism in the brain in intelligent men when they were solving a mathematical puzzle. Higher calorie consumption has historically been an evolutionary negative...
Our extended helpless infancy appears to be necessary for the development of human intelligence. Evolutionarily, extending this period has obvious costs.
It's also possible that creating highly intelligent people is just an extraordinarily hard problem for Mother Nature to solve, requiring a witches' brew of heterozygous recessives and specific conditions for expression. The wide variation in intelligence within each and every human strain suggests to me that this is the case.

Posted by: bbartlog at April 18, 2003 02:32 PM


bbartlog - I didn't say that, I quoted it. Michael Vassar said that (and I should have attributed it in my response). Your second and third ideas as to potential evolutionary costs of higher intelligence sound worth investigating.

godless - maybe engineers learn different stats than life scientists.

What you're saying is that with your normalized variables and r = 0.6 (for example), the most probable value of U (normalized Y) when Z (normalized X) equals 1 is 0.6. But that's not necessarily true. Trivially so - it is conceivable that the most probable values of U when Z = 1 is greater than 1. But r cannot be greater than 1. Therefore r is not a measurement of the slope of the predictor line.

Jason says that the correlation between IQs of parents is 0.45. That would imply, using your definition, that for a person of +2 sigma intelligence, the most probable IQ of their partner is +0.9 sigma. By your definition, that correlation should change depending whether male IQ or female IQ was the X-axis.

By the definition I'm using, for a person of +2 sigma intelligence, the most probable IQ of their partner is +2 sigma, but the range is broader.

r is a measurement of the fit of the predictor line.

Posted by: Anthony at April 18, 2003 05:10 PM


"It's also possible that creating highly intelligent people is just an extraordinarily hard problem for Mother Nature to solve, requiring a witches' brew of heterozygous recessives and specific conditions for expression."

I'm a computer science guy, not a genetics guy, so if i get things wrong, please correct me. I remember reading in The G factor that inbreeding depression causes a drop in IQ scores and Heterosis causes a slight increase in IQ scores and the reason has to do with recessives almost always being deleterious.

".....This is why we find that most disadvantagous mutants in the gene pool turn out to be receissive." (p. 191, The G Factor; Jensen, A.)

"...and nearly all of the seriously deleterious alleles that all normal persons carry are recessive." (p. 194-195, The G Factor; Jensen, A.)

"...Without exception, all of the studies show inbreeding depression both of IQ and of IQ -correlated variables such as scholastic achievement.....the g factor significantly predicts the degree to which performance on various mental tests is affected by inbreeding depression, a theoretically predictable effect for traits that manifest genetic dominance." (p. 194-195, The G Factor; Jensen, A.).

Is there a reason that heterozygous recessives (Aa) increase intelligence on average more than homozygous(AA)alleles? I ask this question because Jensen states on p. 192:

"...With complete dominance, however, the homozygous AA and the heterozygous Aa are phenotypically indistinguishable."

BTW - i find genetics much more interesting than computer science.

Posted by: the alpha male at April 18, 2003 05:23 PM


this shit is why i missed your evil presence....

Posted by: razib at April 19, 2003 02:59 PM


I didn't read all the previous comments: too much text to read. As a member of high-IQ Societies myself, I only want to say this: IQ as well as the 'g' factor (an insufficiently precise concept) is only indicative of intelligence. Intelligence is very much more complex and 'extended' than mere IQ. Besides that, the power-tests of the so-called 'ultra-high-IQ'-societies are notoriously unreliable. There is a simple reason to that: there are too few test cases to give reliable results. For one tests to be normalised in the ultra-high-IQ-ranges (let's say 150+) it would take thousands and thousands of unselected people to undergo the test; for IQs > 180, that would even increase to millions of people. Practically, such extremely extensive tests are never done. Ergo: the so-called ultra-high-IQ societies are mainly clubs of self-flattery, societies of vanity.
Cheers
Jan

Posted by: Jan Snauwaert at July 4, 2003 12:25 PM


I think the intelligence goes beyond what IQ tests can measure, that's all the have to say

Posted by: Alan at August 2, 2003 01:17 AM


I'm sorry, typo on the previous message. I meant
I think that intelligence goes beyond what IQ tests can measure, that's all I have to say

Posted by: Alan at August 2, 2003 01:19 AM


O.K. I agree that intelligence is goes beyond what I.Q tests can measure. The I.Q. tests as I have always known them to be are only indicator of potential.

Off the subject though I have tested at 165 and work currently in the field of computing. I have wondered for some time if there are organizations that recruit specifically based on above average I.Q. scores.

Posted by: Vic H. at October 9, 2003 12:41 AM