Aug 222006

Possibly the most annoying truth in the world is that good qualities cluster. People who are good at something tend to be good at many things.

The psychometric version of clustering is g, or general intelligence. The existence of g is not seriously disputed in the field, and g-deniers like Howard Gardner, with his theory of “multiple intelligences,” or the late paleontologist-cum-Marxist Stephen Jay Gould, are regarded as cranks or axe-grinders. The Wechsler scale, the most complex (and expensive) of all intelligence tests, measures thirteen apparently widely disparate skills, including vocabulary, picture completion, matrix reasoning, and the ability to repeat back a string of digits. Bad news! Each skill, without exception, correlates positively with all the others. Seventy-eight possible correlations, and every one positive. The correlations range from 0.3 (weakish) to 0.8 (very strong), where 0 indicates no relationship and 1 an exact match. If you suck at math, the parsimonious explanation isn’t that your true talents lie in writing or painting. The parsimonious explanation is that you suck.1

Intelligence isn’t everything, you say. Quite so. Nevertheless intelligence correlates quite highly with other happy outcomes, like money and status (0.5 or so), long, healthy lives, and staying out of jail. Not to mention height. About the only negative quality definitely associated with high IQ is myopia.

The news gets worse, much worse. Smart people tend to be good-looking. Since intelligence also correlates significantly with income and status, this should surprise no one. Rich, high-status men marry better-looking women and sire better-looking offspring. QED. Models aren’t stupid after all.2

Jocks aren’t stupid either. Reaction time, essential in many sports, correlates moderately with IQ. The data is spotty for most sports, but pro football players have a significantly higher average IQ than the population at large.3 Again, no surprise; learning an NFL playbook is probably pretty demanding.

Clustering offends the human sense of fairness. The world is frequently imagined as if God gives out qualities one at a time — brains for you, beauty for her, and for him, let’s see, how about a sense of humor? — seeing to it that things somehow even out in the end. Jocks and models must be dumb; they already expended their early-round draft picks on something else. Geniuses are crazy; it’s their handicap. Ugly girls have nice personalities. And most of all, you can be anything you want to be, in Lake Wobegon, where all the children are above average.

1By “you” I mean an abstract individual, a statistical creation, not you, dear reader. No statistics could possibly apply to you. You are special, and Jesus loves you.

2Again, these remarks should not be construed as a reflection on any particular model, like, say, Claudia Schiffer (Schiffer-Brains, as she is known in the industry).

3Assuming the same racial composition.

Update: Matt McIntosh comments. Elsewhere, I mean. The Mechanical Eye comments. I will reply if I can find the time in my busy schedule of feeding Christians to lions, building palaces, and oppressing the peasantry. Ilkka Kokkarinen comments. Degrees of Freedom comments. Noneuklid comments.

  35 Responses to “Clusterfucked”

  1. Ah, there’s that buzzkill I know and love. While not disputing a word in this post, for balance I would point out that people also tend to undervalue the role of brute luck. IQ correlates with socioeconomic status, but only over the long run, and the correlation drops off at the upper levels. Warren Buffet and Bill Gates level wealth is not remotely explicable by their intelligence; stochasticity rules the roost at those heights.

  2. If positive traits cluster, doesn’t that mean, ipso facto, that negative traits also cluster?

  3. Matt: All correlations tend to break down at the tails of the distribution anyway. They’re called outliers for a reason. You are perfectly correct about randomness, but it is a topic for another day.

    Bill: Sure. If smart people tend to be rich, then stupid people tend to be poor. Ipso facto. What’s your point?

  4. “People who are good at something tend to be good at many things. People who are bad at something — not so much.”

    I was questioning the second sentence above.

  5. As I understand it, the existence of general g means more than that many positive attributes are correlated with IQ, it means that some mental attribute, general g, underlies all of them, i.e., the more general g you have, the better you will be at math, music, empathy, etc. The correlations are indisputable, but as I read Gardner and Gould (both cranks, I agree) they dispute the existence of some general mental attribute that underlies all the other abilities.

    Gardner and Gould are right. The existence of rare individuals who have a great deal of some ability, like calculating prime numbers or composing music, while having sub-normal scores on IQ tests or even being unable to care for themselves (we used to call them “idiots savants” in less sensitive times) shows that general g, as a separate mental attribute, does not exist.

  6. Aaron’s father (Dad, that would be): The existence of savants doesn’t disprove the existence of g. They’re the outliers Aaron referred to above, with one clearly extraordinary ability and some equally extraordinary deficits. For the vast majority of us, who aren’t extraordinary in any way, g can explain why ordinary talents cluster.

  7. Without disputing your claim that g corellates positively with income, I believe Steve Sailer recently said that the correlation coefficient was something like .l0 nor .50, which is a big difference. His point was that g counts but a lot of other stuff does too. Maybe I am remembering him wrong, but would you care to document this particular figure of .50? thanks

  8. Aaron: Yup.

    Aaron’s dad: This is a rare moment — you are, for once, way wrong. We’re already well on our way toward figuring out the neurobiology of g. Start here and here. Think of all the skills that go into an IQ score as different software modules and g as the hardware they run on.

  9. Here is the exact quote I referenced above:

    “Consider IQ. An important thing to keep in mind is that when measured at the individual level, the importance of IQ, while worth studying, is limited. Human behavior is hugely complicated, and any one measure can explain only a small part of it. Thus the r-squared correlation between IQ and income is somewhere in the range of 10%. That’s actually a pretty large number relative to other measurable factors, but obviously the glass is only 10% full and 90% empty here.

    On the other hand, when looking at two groups that differ in average IQ, IQ is hugely important. For example, if you look at the graduates of the 20 highest IQ colleges in the country and compare them to the graduates of the 20 lowest IQ colleges in the country on current income, I would bet that at least 19 of the higher IQ colleges, and probably all 20, came out above the lower IQ colleges in income per graduate.”

  10. Sorry Luke, I crossed the links, which are now fixed. I’m quoting Alex Tabarrok on 0.5 (though I have no idea how one would measure “status”). The Gene Expression boys say it’s 0.33. Neither cites a study and casual Googling doesn’t turn up any good ones. The Bell Curve cannot produce a number; it relies on a highly inferential analysis of the “residual” that exists in income distribution after correcting for education. So until further notice I think we have to regard the 0.5 figure as suspect.

  11. Aaron,

    Yes, g most likely exists, and a good brain typically adapts and performs well no matter what the task. But this does not tell the whole story.

    Cognitive ability is just that. It relies on the brain’s ability to process information, store the information, categorize and retrieve the information, and so forth.

    All of these processes rely on the strength, density, and architecture of the brain’s neurons, both in micro- and macro architecture. It is the efficiency of these connections that make for “brain power”, think of the “noiseless channel” ala Shannon.

    Take schizophrenia or autism as examples, these disorders are likely failures of brain connectivity, without this necessary feature, brain function becomes disorded.

    Once one has adequate brain power, one must then use it for adapative purposes. This crucial step requires additional variables such as one’s environmental and monetary resources, social connectivity, mentoring, etc.

    Without g life is often hard with limited “success” no matter what the metric.

    But again, it is in the “choosing wisely” along with rich resources that will most likely confer success.

  12. MeeTooThen: What, and you think “choosing wisely” doesn’t correlate strongly with g as well?

    BTW, since this is your area I was wondering if you’d ever read this book. I was considering plunking it onto my reading list but am concerned it would be above my level.

    (As an aside, I don’t understand why phrases like “[g] does not tell the whole story” keep cropping up in this sort of discussion. Has anyone ever asserted the converse? It’s like a sort of strawman cum throat-clearing.)

  13. Matt,

    Thank you for the kind reply.

    No, from personal and professional experience choosing wisely does not necessarily correlate with g.

    Human nature and all that.

    And yes, perhaps I was a bit clumsy in my comments (what else is new?).

    My point was to add to the notion that g is a mostly necessary but not sufficient attribute for success.

    As to exploring the neural code, in years past (now going on 20 years) this was the kind of topic that was being talked about but not yet widely explored.

    I recall hearing for the first time in 1992 chaos modeling of interictal epileptiform spikes as a way of understanding epilepsy and its neural substrates.

    Thanks for the heads-up. I will order the book and read it straight away.

  14. Aaron,

    From Scientific American, more on g, here.

    Just sayin’.

  15. MeTooThen: Cool, let me know how the book is. You’re right about g being necessary but not sufficient, which is why I would still disagree with you on the other point — the correlation between g and wise choices might not be strong, but I would bet an arm and a leg that it is positive.

    And Linda Gottfriedson is interesting on this subject in general. I have not yet read many of her papers, but you can find them all here.

  16. Matt, the existence of a specific brain location for general g would, of course, clinch the argument for its existence, but the references you give make no such claim. There are specific but distinct brain locations for recognizing faces and putting a name to them, for example, but this is evidence against the existence general g. If someone finds a brain location which, on suffering damage, diminishes most of the abilities correlated with general g, that would establish g’s existence, but I don’t believe this has happened. Of course, g might exist without having a specific brain location, but this claim also requires proof beyond correlations.

    Tamar (Aaron’s sister), it’s good to see my kids ganging up on me, it makes me feel younger. If g exists, then it exists in all cases, not just in most, or all but some outliers, and nobody can have a great deal of ability in some area and be sub-normal in all the others which supposedly also depend on g. The exception proves the rule (I have been waiting years to use this phrase correctly.) and in this case the rule is found wanting.

  17. Aaron’s Father,

    The neural substrate of g is the brain, or most of it anyway.

    In other words, there will not likely be a single or discreet locus for g just as there isn’t a discreet locus for consciousness.

    Our understanding of the neuroanatomical localization of brain function has changed to the point where given a well known and easy to describe function such as motor language, (and its substrate, the so-called Broca’s area) it is now believed that motor language is subtended over a more widely distributed region within the same brain, and over a more diverse pattern amongst differing individuals’ brains than was previously thought.

    While it’s true that single brain lesions can destroy consciousness, vision, motor strength, etc., for such complex brain behaviors such as consciousness and g, their neural substrates are likely widely distributed over both hemispheres and involve cortical, subcortical, and brainstem regions working simultaneously (or contemporaneously) as a dynamical system.

    Again, g reflects (in part) the brain’s overall power, its efficiency of processing and retrieval of information, and its connectivity both at the micro- and macro-levels.

    Having said that, there does likely exist differing “weighting” within individual brains, or put another way, heirarchies of brain power within the brains subsystems that allows for an individual to be a great writer but poor mathematician. Hence g is representative but not fully predictive of brain function, or ability as distinct from intelligence.

    I believe it is this inter-subsystem variability that turns folks off to the notion of g.

    But to my mind anyway, the presence of differing abilities (manifest behaviors) seems a natural consequence of a complex system and its constituent subsystems.

    Just sayin’.

  18. Aaron’s dad: You’re making a category mistake. g is not a physical thing you could touch, though it is based on a physcal substrate as MTT says — it’s an abstraction, and qua abstraction, g does exist unquestionably. If you want to say otherwise you have a hell of a lot of very consistent data to explain away. We’re still exploring the ways in which it’s defined by the physical substrate, but we don’t need to understand that in order to know that g exists. It’s much like a “self” or a center of gravity — there is no physical part of a body you can point to and say “this is a self” or “this is a center of gravity”, but that doesn’t make them any less real and useful abstractions.

  19. Let’s get off abstractions and onto particulars. Do the tests performed on Ruth and Pujols here suggest something about that weird letter not on MY keyboard?

    What about Pujols’ eye movements?

    Clearly Ruth and Pujols have good hardware, but do ya think either could understand relativity? Alpha theory?

    (I don’t have any doubt that Wiliams could have.)

  20. “The news gets worse, much worse. Smart people tend to be good-looking.”

    I am willing to bet this correlation is only due to recent breeding patterns of the last couple hundred years of the West, starting maybe about a hundred years before industrialisation when intelligence was first systematically culled and brought into larger professions with higher standing.

    Back in the day, intelligence could provide some luck and additional survivability, but the poverty of material conditions meant major and generally random upheavals and setbacks for the intelligent “class”, which made for generally random breeding. Only recently have intelligent people begun breeding in stratified groups with short ranges. Since women fuck “upwards” socially* and men fuck partly neutrally -but heavily weighted toward beauty- the fairly common occurance of beauty would quickly become correlated with the fairly uncommon incidence of brains.

    *- This is an obvious social trait of women, but it’s also been proven statistically: women tend to marry high IQ men more than low-IQ men, with the higher IQ, the greater likelyhood of marriage and breeding. It’s probably the only thing keeping us from beeding out intelligence entirely. Men don’t seem to go after intelligent women the same way they go after beauty, and and that’s combined with the intelligent women being pickier than “normal” women. Insert sneering feminist comment here_______________________. Thank you.

  21. “Jocks aren’t stupid either. Reaction time, essential in many sports, correlates moderately with IQ. ”

    You don’t understand this one correctly. Reaction time isn’t the correlate of athleticism. Athleticism is correlated to the speed, fluidity and efficiency of muscle response ONCE THE RESPONSE HAS BEEN MADE*. Reaction time among whites is significantly faster than blacks, but the completion time of large or complicated athletic movements is much faster among blacks. It’s the difference between reaction time and completion time.

    Reaction time is not correlated with athlecticism, it’s correlated with the athlete himself. The (quite mild) correlation with athletes and IQ is instead a function of both the higher concentration and better long-term training habits of the more intelligent athlete (thus giving more long term return and fuller “realization” of any given athletic ability). That’s also why the correlation is so mild: higher IQ is not a cause, it’s an effect.

    *- Apologies for the caps.

  22. Sorry for the multiple postings- I just keep thinking of more crap to say.

    “Clustering offends the human sense of fairness.”

    Only in men, I think. I don’t think women naturally field egalitarian sentiments. Both men and women have their unspoken social and biological darwinisms, but women I’ve noticed seem less forgiving of weakness and more admiring of strength than men are. Again, this is the largest generality, but women just seem biologically OK with hierarchy. Or, rather, men’s job is to place himself well in the hierarchy, and women’s job is to judge the man’s place.

    I’ve often seen women more unforgiving of unfair natural circumstances than men would ever be. Were it not for women’s natural ironic compassion (needed to foster children and avoid being killed by the larger and more violent male), they would be savagely darwinian, as they always are in unguarded moments. I’ve never seen a woman who was naturally concerned with abstract, general notions of fairness or justice. Whenever it came up, there was always some other tangential reason that was clearly the driving force. Fairness has always been the man’s preoccupation.

    I did an argument with an “alternate psychometrician” once, one of the Penn State alty/Gardner types who “disagreed” with g. What always gets me is how addlepated the answers were to the basic question: “Why are scores correlated so heavily across fields?”. Well, er, you see, not all scores correlate exactly as a..(trails off, mumbles). “That’s why they call it g. It stands for general”. Well, um, some people are good at some stuff, and, uhhh. (fades out). Wanna go to McDonalds?

    I’ve never gotten a concise answer to the most basic question, a clear indicator something is wrong. In the sciences, it’s a no-brainer: it’s proof is in the discovery. Either it correlates generally or it doesn’t. If it does, g is proved.

    I am confused about it, though: it’s counterintuitive with what we know about the brain (this minute, anyway). Apparently the brain is a bunch of specialized nodules that differentiates specifically in individual brains. It seems like that would be much more specialized than it is. Go figger.

  23. James, this puzzled me a bit as well but I realized that modularity and g are not in conflict. It’s why I used the hardware/software analogy above: having more RAM and a faster processor will improve the performance of all the software on your machine across the board, even if each of these modules does very different things.

  24. Best or the Worst?

    GOD OF THE Machine makes a dispiriting assertion – that is, the more intelligent you are, the better looking you are, and vice versa: Possibly the most annoying truth in the world is that good qualities cluster. People who are…

  25. I didn’t read any of the comments before responding, so I’m not quite sure what you’re talking about. I’ll have to go over and look at what you’d said to understand what you mean. I also am not sure which posting you responded to.

  26. Oh ok, I see what you’re getting at now.

    I dunno. I tend to think we’re not at the point yet where we can account for g alongside savant abilities: some have suggested the unused pathways in parts of the brain of low-IQ people may leave the brain more unused pathways for what isn’t being stopped by “poor wiring”, thus giving idiot-savants another edge. Somewhat similar to yours, but also different. From my understanding of neurobiology, the idea of the “module” brain is completely in its infancy, and amounts to a guess as to the general structures.

    But g is hard to supress as an idea about the brain’s functioning: it’s very applicability in so much of the population suggests the explanation for savants is going to involve some kind of unique, individual tuning for their brains, and not some other whole areas of competancy as Gardner so fanatically hopes. Overall though, I think it’s way too early to make guesses like the ones we’re making: me with modularity and you with the computer dealie don’t even rise to the level of educated guesses at this point. g…not so much.

  27. “I am confused about it, though: it’s counterintuitive with what we know about the brain (this minute, anyway). Apparently the brain is a bunch of specialized nodules that differentiates specifically in individual brains. It seems like that would be much more specialized than it is.”


    My own view of the paradox is that there may be some underlying substrate, mechanism or condition that accounts for it. For example, different abilities in the various specialty nodes may be overshadowed by, let’s say, large glial cells throughout the brain.

    Your description of women is unique and similar to a view I have sometimes held.

  28. James,

    I disagree. We must create the hypotheses for falsification.

  29. Well that ain’t a structural description, is it?

  30. Wait! This just out–tall people are smarter.

  31. I like how it wasn’t presented as a simple fact, but immediately used to justify a previously known bias in favor of tall people!

    And again, we find that people seem to really care which couple of letters on the genome they can discriminate against- the cluster of race? Sacre bleu! But the portion that deals with height, weight, chronic halitosis or eyesite? Have at!

    So, when a woman says “I want to be appreciated for my mind!”, just ask her to map out the portion of the genome she wants to be appreciated for and which bits are off limits.

    I am constantly reminded of something my grandfather once said: “You know, they used to say people with discriminating tastes were good”.

  32. Chances are…

    Over at Catallarchy there is an entry titled “How do you make a leftist’s head explode?” …. pause to see if reader is still with me … and the dude makes a point or two. The link to “D” from…

  33. It seems to me that people who are not cultists of g have a much higher putative g than those who are (Gould, for example, compared to say Charles Murray), which seems significant somehow. There’s also the existence of Mensa.

    If that’s not enough for you, take a look at the Errol Morris First Person interviews with Rick Rosner and Christopher Langan.

  34. Your faith in ad hominem (if believers in g are “cultists,” then mainstream psychometrics must be a cult) and anecdote is touching. Gould never explains the correlations among the various skills, and neither do you. g is a parsimonious explanation. Have you got a better one?

  35. Thomas dit :Je ne vas pas plus loin que les chiffres et les conclusion que l’étude tire.La lecture du document démontre que l’étude ne tire aucune conclusion. L’étude dresse des profils de la consommation de produits culturels au Cae;ra.C&dsquoanst descriptif et aucune inférence n’est faite.

 Leave a Reply

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>