April 6, 2013

Is the g Factor a myth?

At Human Varieties:
Is Psychometric g a Myth? 
Posted by Dalliard

As an online discussion about IQ or general intelligence grows longer, the probability of someone linking to statistician Cosma Shalizi’s essay g, a Statistical Myth approaches 1. Usually the link is accompanied by an assertion to the effect that Shalizi offers a definitive refutation of the concept of general mental ability, or psychometric g. 
In this post, I will show that Shalizi’s case against g appears strong only because he misstates several key facts and because he omits all the best evidence that the other side has offered in support of g.

Shalizi has a lively intelligence, but he has a little too much confidence in his own g Factor. He begins his essay:
Attention Conservation Notice: About 11,000 words on the triviality of finding that positively correlated variables are all correlated with a linear combination of each other, and why this becomes no more profound when the variables are scores on intelligence tests. ... To summarize what follows below ..., the case for g rests on a statistical technique, factor analysis, which works solely on correlations between tests. Factor analysis is handy for summarizing data, but can't tell us where the correlations came from; it always says that there is a general factor whenever there are only positive correlations. 

But why are there only positive correlations among cognitive skills? That's hardly a trivial question. Many things in life come with tradeoffs, such as risk v. reward. 

Shalizi gives an example about cars, in which he misses this crucial point:
One of the examples in my data-mining class is to take a ten-dimensional data set about the attributes of different models of cars, and boil it down to two factors which, together, describe 83 percent of the variance across automobiles. [6] The leading factor, the automotive equivalent of g, is positively correlated with everything (price, engine size, passengers, length, wheelbase, weight, width, horsepower, turning radius) except gas mileage. It basically says whether the car is bigger or smaller than average. The second factor, which I picked to be uncorrelated with the first, is most positively correlated with price and horsepower, and negatively with the number of passengers — the sports-car/mini-van axis.
In this case, the analysis makes up some variables which aren't too implausible-sounding, given our background knowledge. Mathematically, however, the first factor is just a weighted sum of the traits, with big positive weights on most variables and a negative weight on gas mileage. That we can make verbal sense of it is, to use a technical term, pure gravy. Really it's all just about redescribing the data.

The first factor is less bigness than an axis of affordability v. something like "impressiveness." If you look closely at the components, you can see the tradeoffs: for example, Shalizi implies that horsepower and price are positively correlated, but it's more insightful to think of them as inversely correlated. Restate "price" as, say, "change left over from a $100,000 bill" and the affordability v. impressiveness trade-off is obvious. This inherent tradeoff is one that automobile engineers and marketers struggle with everyday.

In contrast, we don't see these the same levels of tradeoffs on cognitive abilities. Sure, autistic savants might have some amazing skills because they lack others. And blind people sometimes are better at, say, music or other tasks involving mentally processing nonvisual inputs. What's surprising is we don't see these tradeoffs as much.

But, with cognitive tasks, on average, we don't see the kind of tradeoffs you see with most engineering problems. People who are above average on math skills are not, on average, below average on verbal skills. On average, they are above average on both. 

That's kind of strange when you think about it. Presumably, there are trade-offs involving, say, head size, ease of childbirth, nutrition needs, hip width and running speed, tendency to fall over, tendency to get abstracted, and so forth. But the lack of overall tradeoffs just within cognitive tasks is pretty odd. 

66 comments:

Assistant Village Idiot said...

They had ANOVA in the 19th C, when IQ was being correlated to abilities outside of the lab altogether? Who knew?

Mr Lomez said...

I understand Shazili as merely saying that a common factor 'g' to explain correlations across cognitive abilities does not HAVE to exist. To which I agree. It's possible to be statistically consistent and say that 'g' is a fabrication sewn from the cloth of the data.

What Shazili ignores, of course, is whether, given this data, 'g' is LIKELY to exist (regardless of whether its exact and specific components can be measured). Unless someone were actively searching for reasons not to "believe" in 'g', the answer is resoundingly in the affirmative.

Seems to me like just another case of Occam's butterknife.

Jim Bowery said...

There's a simple-enough Monte Carlo test of whether the data-mining fallacy is plausibly applied in a particular case:

Generate random data of the same general characteristics (dimensionality, major statistical moments like mean, variance, skewness, kurtosis, etc) and see if the factor analysis shows approximately as good explanatory power. If it doesn't then one has to ask one's self how likely is it that the actual data's factor analysis arose by chance.

Brett_McS said...

There is no trade-off within software: Good software runs on the same hardware as does bad - just with a better result. Similarly with good recipes: Same ingredients, different outcome.

The British astronomer Fred Hoyle proposed that ability with maths was more about the way it was learnt in a person's early years. There are good ways, there are bad ways.

Bill said...

Well, it seems that there is a pretty big tradeoff in a most important area:

http://en.wikipedia.org/wiki/File:Countriesbyfertilityrate.svg

Charlesz Martel said...

I don't understand your "inversely correlated" comment. Please expand.

realist said...

Do we really know all cognitive abilities are positively correlated? How many cognitive abilities have we really studied? How many cognitive abilities exist?

Maxwell said...

I think the official word (as of a 2009 study) is that the existence of the g factor is "unproven and subjective, based on a questionaire".

That being said, there have been many studies on the g factor for 70 years now, and ultrasound studies have found significant physiological evidence for its existence.

What is known for certain is that it is not usual for women to recognize that they can experience their g factor, but strangely enough there is actually a surgery that can amplify its effect.

Misophile said...

Is it impossible to be both a race realist and believe "g" is a nearly meaningless statistical artifact? If you're interested in what intelligence actually is, on a theoretical level, or (say) if you want to build an AI, g tells you absolutely nothing. Shalizi is more-or-less right, but at the same time you CAN'T argue with predictive power. You can make predictions with g, but you can't make anything else because g is an empty concept.

Cail Corishev said...

It would seem odder if there were clear trade-offs between different cognitive abilities. If your brain is good at processing one kind of information, accessing memory, making connections, and so on, it'll probably be good at others.

This is mostly true for athletes too. On the margins, a super-tall guy might excel at basketball but be bad at soccer or wrestling. But for the most part, the kid who excels at kickball in first grade will also be the best at dodgeball, baseball, basketball, and other sports as he comes up through school. At smaller schools, it's not unusual for the star quarterback to also be the star of the basketball team and the baseball team's starting pitcher and clean-up hitter. Agility and muscle memory apply well across the board, just as cognitive "agility" applies well across mental tasks.

Going back to academics, again, in small grade schools, the smartest kid is usually the smartest at everything: he's the best at math, reads ahead of everyone else, wins the spelling bees, memorizes best in every subject, and even learns a musical instrument the fastest (even if he's not the most musically talented). g is so obvious in that setting that it's hard to believe people seriously argue against it.

It's only in schools large enough to have savants show up in particular subjects or sports, or later in life once people have started to specialize, that you can not notice that people who are really good at one thing tend to be good at related things.

candid_observer said...

Shalizi's arguments exhibit the sort of arrogance only complete ignorance can render possible.

Does he really think that it has occurred to no one in the history of psychometrics that g might be some kind of artifact, or that that question hasn't been addressed extensively by many, many smart people -- including some, such as Spearman, who were among the founders of statistics? How can Shalizi possibly believe that such a field, as it was developed, would countenance in its most basic concepts howlers like those he describes?

I can't help but think that he took his cue from his fellow charlatan, Stephen Jay Gould, who likewise claimed that the simplest argument against g had never dawned on anyone in psychometrics.

Apparently, Shalizi no longer wants to talk about IQ and this issue. I think we can surmise the answer as to why: because he's been owned repeatedly by those not in his throes of ignorance.

Cail Corishev said...

The myth is that being good at one kind of thinking means you're probably bad at another. Usually the division is between math and everything else, but some people also like to draw the line between book smarts and common sense. We like to think that because it makes things seem fairer.

This is probably because most people had a subject in school that they enjoyed most of all, so they think of that as their "good" subject and the one they enjoyed the least as the one they're "bad at." I thought of myself as good at math and bad at history -- but I was a straight-A student. The truth is I was good at all my classes; I just enjoyed math a lot more than memorizing names and dates in history class. That's closer to reality -- people who are good at one cognitive skill are usually good at others, if not at the exact same level on each.

What g tells us is that the mental process that works out a math problem isn't as different from the process that figures out the meaning of a poem as we'd like to think. The brain isn't like a screwdriver, designed for either Phillips or straight-edge screws, and useless for the other. It's more flexible than that, and most mental problems in any subject come down to the same skills: reason, memory, logic, and making connections.

The amount that g doesn't explain is probably the amount that one problem, like math, takes a different ratio of those skills (high on logic, for instance) than a problem like winning a spelling bee (higher on memory). But then you come back to the fact that a healthy, well-working brain that's more able than most will probably do both better than average, just like the athlete with agile feet will usually have better-than-average hand-eye coordination too.

Luke Lea said...

"Shalizi has a lively intelligence, but he has a little too much confidence in his own IQ."

As do his followers. They point to him (as he points to himself) as THE AUTHORITY. I once suggested he take it up with the American Statistical Association instead of the man in the street.

Anonymous said...

Ross Douthat is stealing your ideas again in Sunday's NYTimes column.

Evan McLaren said...

The "common knowledge" that IQ is a bogus measurement and instrument has seeped down pretty deep, down to people I meet everyday who I know are not as tortured by the debate as I am.

Anonymous said...

the g spot is a myth. g factor is real.

Five Daarstens said...

"The British astronomer Fred Hoyle proposed that ability with maths was more about the way it was learnt in a person's early years. There are good ways, there are bad ways."

Developing software is not only about maths, but about language and music skills as well. Many of the early software developers were very much into music.

Anonymous said...

tradeoffs in mental abilities:

Chimps are stupid but they are better at memorization than humans are.

http://www.youtube.com/watch?v=DqoImw2ZWmI&feature=player_embedded

Aaron Gross said...

I've only skimmed the article so far, but the most obvious gap in Shalizi's argument is that he totally ignores neurophysiological correlates of g. "So far as I can tell, however, nobody has presented a case for g apart from thoroughly invalid arguments from factor analysis; that is, the myth." Jensen spent a good part of the 1990s presenting just such a case.

To me, the most interesting thing in Shalizi's article wasn't about g at all, it was this: "[The Big Five personality model] quite robustly fails confirmatory factor analyses: the `Big Five', despite being made up for the purpose, don't actually fit the correlations in the data, even on personality tests designed using the theory." [emphasis in the original]

I haven't followed up on this. Is it true?

Steve Sailer said...

"But for the most part, the kid who excels at kickball in first grade will also be the best at dodgeball, baseball, basketball, and other sports as he comes up through school."

Right, the kid who totally dominated first grade kickball in my class went on to hit 49 homers in the minor leagues of baseball over five seasons.

Steve Sailer said...

Is it impossible to be both a race realist and believe "g" is a nearly meaningless statistical artifact?"

Sure.

Imagine that upper body strength was totally uncorrelated with lower body strength, so there's no g factor for strength. And, say, that blacks averaged 20% more lower body strength and whites 10% more upper body strength. Then you'd see whites dominate weightlifting and blacks dominate sprinting. In football, where both count, you'd see disproportionately more blacks than whites.

Now, racially, that doesn't look too different from the real world. But in this world, there sort of is a g factor for strength.

Steve Sailer said...

Conversely, in "The g Factor," Arthur Jensen argued that men and women were equal in g.

Really, when you stop to think about it, empirical equality is highly unlikely without g factors. If there is just one big factor, it's not all that implausible that two groups of people could be about equal on one number. But if there are, say, 8 factors that all matter significantly, then it's highly implausible that two groups happen to be identical on all 8. I pointed this out to Howard Gardner and he agreed with me.

Anonymous said...

Ultimately statistics can only tell you something about where to look for likely causes.

If we observe a g we'd expect to find an underlying physical phenomenon like neuro-conducting velocity, short term working memory size, or the like. The statistics can give you clues about where to look, but there's a difference between "we observe g" and "certain people have a higher IQ because they have faster nerve conduction velocity." It's the difference between "this coin comes up heads 75% of the time" and "This coin is weighted on the heads side which makes it more likely to come up heads."

Anonymous said...

g is probably negatively correlated against development/maturation time, cost, percentage of heatstrokes along the equator, and so on. Just because we used g to mitigate/eliminate all that doesn't mean it isn't there.

jody said...

"But, with cognitive tasks, on average, we don't see the kind of tradeoffs you see with most engineering problems. People who are above average on math skills are not, on average, below average on verbal skills. On average, they are above average on both."

all the research i'm familiar with, as well as all my real world anecdotal experience, says this is generally true.

however, check out this spectacular exception:

The Professor, the Bikini Model and the Suitcase Full of Trouble

(new york times link) http://tinyurl.com/ae9x7cj

tenured physics professor at UNC, who has co-authored papers with nobel physics prize winners, is tricked into believing he is talking to denise milani on the internet, and agrees to meet her in bolivia. it gets worse from there.

i refuse to believe anybody on earth is actually that stupid. yet, there he is.

jody said...

"This is mostly true for athletes too. On the margins, a super-tall guy might excel at basketball but be bad at soccer or wrestling. But for the most part, the kid who excels at kickball in first grade will also be the best at dodgeball, baseball, basketball, and other sports as he comes up through school."

true. "g" eneral athletic ability will get you to the top of several sports at the high school level. you have to start to pick at the D1 level though. by pro there's only a couple guys left who can do more than 1 sport.

US high school sports participation is up over 4 million per year now, so the sports world is a steep funnel from the bottom to the top. you go from 4 million teenagers playing about 10 sports, with thousands of them playing several sports at the same time very well, and 4 years later, by the end of college, you're down to a few hundred guys going pro in 1 single sport and maybe only a handful who still have to choose between 2 sports.

Anonymous said...

Perhaps the best evidence for the reality of 'g' is the correlation to the 'reverse digit span' test, and the direct evidence of measuring evoked electrical potential 'brain waves' through sensors placed on the scalp.

Anonymous said...

And blind people sometimes are better at, say, music or other tasks involving mentally processing nonvisual inputs. What's surprising is we don't see these tradeoffs as much.

The reason for this is that what's happening mechanically is that the brain is co-opting a visual input center and re-purposing it for another kind of cognitive ability.
Sometimes this works out, but often what happens is that, with all the extra problems of connectivity (visual areas are relatively far away from audio processing regions) and re-purposing a region which, while relatively undifferentiated, does have an evolved function, often persons who are blind work out not better off.

Autistic savants are not quite the same, in that they typically, neurologically, have large brains but with awkward connectivity (either more or less or weirdly arranged compared to the typical brain) so their brain's neurological activity gets biased towards what most neurologically normal people would call obsessing over certain "irrelevant" stimuli.

However, looking at "irrelevant" information with a lot of cognitive processing (brain size is one contributor to this, even if a weak one - the within population slopes on g and brain size are pretty unimpressive, even though the correlation coefficient is high) sometimes gives you interesting results!

sunbeam said...

Someone mentioned in a previous thread, and had a link to some machine that supposedly could directly measure the intelligence of an individual in a few minutes.

Supposedly there was a pretty good correlation between measured intelligence and the machine's results.

Now to me this sounded like something John Campbell would have pitched back in the day after a drinking session with old L. Ron.

But why couldn't you do something like this? I mean if genetics has a direct affect on intelligence, then the brain of an intelligent individual is going to have a different structure in some way from that of one that is not considered intelligent.

And there is no way to measure this difference? I'm no neuroscientist, but it sure seems likely to me you could do something with the idea.

Anonymous said...

One of the great yacht designers said to a customer who had presented his laundry list of features he wanted in his next custom boat, (paraphrased) "A boat can be fast, comfortable, inexpensive. Pick any two."

Ray Sawhill said...

Hey, I notice that many of you are making a distinction between cognitive-type G powers and athletic talent. (Cognitive G vs athletic G, maybe.) Very useful and very interesting. Plus it gives me a chance to return to one of my hobbyhorses: artistic talent is more akin to athletic talent than it is to cognitive oomph.

It's a myth (promoted by people who are generally pretty high G themselves, and who like to think that the artists they study and promote are smarter than they often are) that art-talent is a cognitive talent. Sometimes having some serious cognitive G can enhance an arty person's accomplishments and career; sometimes it doesn't. But the basic talents that arty people possess have (as far as I've been able to tell in 35 years of hanging around various creative fields) almost nothing to do with IQ.

Incidentally: yes, yes, everyone can point out a bunch of creative types who were clearly smart people. Just keep in mind that it's equally as easy to point out a bunch of creative types who were/are pretty dim in IQ terms.

It's interesting too that while, as everyone here is noticing, in a small school the brightest kid is likely to be one of the brightest kids in every academic class, and the most athletically gifted kid will often be one of the best kids at every sport, artistically talented kids don't tend to sort out like that. It's rare for the kid with a knack for visual style, for instance, to also be one of the school's best musicians; or for a kid who's a natural-born actor to also show a lot of photography talent. (I do know some people who are unstoppably creative in a kind of abstract way -- they do it all, and they seem to have a knack for it all. But they're pretty rare. Usually, you're either a performance person, or a visual person, or a music person, or a words/stories/poetry person ...)

And before anyone brings up Great Art: these *are* the kids who are likely to go on and be active in a cultural-creative sense: as musicians, designers, writers, producers, etc. Besides, if we can discuss the kids who are at the top of a small school's academic ladder and/or athletic ladder as though that shows us something about how the world really works, why can't we similarly consider the kids who are at the top of a small school's creativity ladder?

And that's all before we start discussing things like drive and luck ...

Ray Sawhill said...

BTW, I don't know why some people around here have such a lot of trouble with the idea that artistic talent is more akin to athletic talent than it is to IQ-style G. If you can wrap your mind around the notion that athletic talent and IQ-style G are two different things ... If you can recall what life was like back in high school, where kids often sort out into jocks and brainiacs ... Then what's the big leap? There was usually (for instance) a theater crowd too, as well as a decent rock band, as well as some oddballs who published a whacky 'zine. (OK, I'm dating myself. Maybe these days the creative kids tend to maintain weirdo Tumblr blogs.) There's often a little overlap between the brainiac, the athletic, and the creative worlds in high school, but not a lot. They have their own identities. Same pattern holds in real life.

Do some people have trouble thinking of creative talent as distinct from IQ-style G just because they've been taught to think of the arts as a subcategory of intellectual achievement? If so: Hey, kids, your teachers and profs have misled you. Yet another reason to mistrust your establishment brainwashing, er, education.

Most of the really bright people in the arts sit in offices. They're editors, producers, administrators. The creative crowd (photographers, poets, designers, etc) is something else entirely.

Anonymous said...

In another article he wrote on IQ, Shalizi writes "Whether IQ means anything or not, it is, unlike general intelligence, unquestionably something we can measure . . ." This seems to open up some problems for anyone who thinks g is a myth.

In the same article's conclusion, he writes, regarding whether or not IQ is heritable and malleable, "My honest answer would be that I presently have no evidence one way or the other. If you put a gun to my head and asked me to guess, and I couldn't tell what answer you wanted to hear, I'd say that my suspicion is that there are [differences of IQ between people], mostly on the strength of analogy to other areas of biology where we know much more."

Here's the article: http://vserver1.cscs.lsa.umich.edu/~crshalizi/weblog/520.html

Handle said...

It doesn't seem so odd to me. Doesn't a computer get better at all the various things it does when it has a better microprocessor? There are physical/survival tradeoff between microprocessor performance in general and size, heat, etc. but not between the various tasks the chip is asked to perform.

Still, I think it is more interesting to note that many exceptionally brilliant people are very smart at most things, but seem to be disproportionately smart (or, just as often, lacking) in one particular cognitive area. Thing of people who have rare gifts at learning foreign languages past youth. Proficiency at certain games like chess also comes to mind. On the other hand, thing of very high level mathematical skills - which many very intelligent people cannot even approach.

Maybe it's something like car engines. There are tradeoffs between engine size and other physical and economic constraints. But all things being equal, a bigger engine can accelerate faster as well as tow larger loads. But some equally bigger engines are specifically designed to make fast sports car while others are tailored for slow but heavy duty hauling trucks.

candid_observer said...

Steve raises a very good point regarding the considerable non-obviousness of the fact that almost all cognitive abilities are positively correlated.

Shalizi's entire argument is really little more than a recap of Gould's: that a factor such as g rises pretty much inevitably in any phenomenon subjected to factor analysis, so why pretend that g is anything special or real? Yet that claim couldn't be more obviously false. In the study of a vast majority of phenomena -- as any statistician should be aware -- the correlations found are all over the map -- positive, negative, and, perhaps most commonly, very near zero. What is surprising in the study of cognitive abilities is that they are all positive, and, relatively speaking, quite high. That g usually explains more than all other principal components combined is especially remarkable.

That Shalizi can pretend this is not an important and otherwise unexpected finding is a testament to his ignorance and wretched judgment. If you were to strip him of his panoply of software, and the exciting buzzwords that attend it, he would be instantly recognized as an idiot.

Anyone with developed scientific instincts would recognize that the existence of g at least raises the serious possibility that there is some underlying mechanism that supports it; correlations don't require a causal explanation, but they do generally require some kind of explanation, and in science those explanations tend to be causes. Perhaps all those positive correlations are some kind of very odd concurrence -- but that is certainly not the most reasonable first hypothesis.

candid_observer said...

In a word, Gould has passed the charlatan torch on to Shalizi, who now runs proudly before us in his stead.

David said...

Imagine a line of cars. Some old Chevy Novas, a '73 Cadillac, a Ford F-150, and a brand-new F1 race car of some kind. Which is fastest? The F1 race car. Ah, but not so fast, you car-ist! There are measurable differences in performance among these vehicles. Imagine that one of the Novas has a faster pick-up for the first fraction of a second when the race begins. Then for .001 seconds, it will be ahead of the F1 race car. Or consider the F-150. It can haul a large amount of weight; the F1 race car can't haul nearly as much (just the weight of the driver plus a little). Perhaps it's the lumber in the bed that's holding the F-150 back in the race, not its engine and aerodynamics. Didja ever think of that? Didja? We can even measure the differences in the air drag between the two vehicles and then compute the, etc. Add up a lot of these trivial differences, and you've got yourself a decent cloud of squid ink... but the F1 race car, ceteris paribus, will win the race. It is the fastest overall.

The squid ink explored above is no more absurd than this: "the case for g [that is, for general intelligence - i.e., the universally observed fact that some people are overall smarter than other people] rests on a statistical technique, factor analysis, which works solely on correlations between tests. Factor analysis is handy for summarizing data, but can't tell us where the correlations came from; it always says that there is a general factor whenever there are only positive correlations. But why are there only positive correlations among cognitive skills? That's hardly a trivial question. Many things in life come with tradeoffs...."

Note the false first premise: "the case for g rests on a statistical technique, factor analysis." This is like when Mr. Burns in The Simpsons began one of his arguments with: "Since the beginning of time, man has yearned to destroy the sun." Factor analysis quantifies general intelligence for study; factor analysis neither discovered it nor is the primary evidence for it. General intelligence was discovered in the caveman days. The evidence for it is ubiquitous. What we (maybe) need is to explain it, not explain it away.

There are people who devote their lives to perceiving and clarifying reality, and there are people who devote their lives to denying and obfuscating reality. The latter may be found anywhere, but in modern times they are especially prevalent in "philosophy" (which used to mean "love of wisdom" but now means destruction). These people thrive on attention, so I recommend giving them only as much as is needed to knock down their nonsense in the shortest way.

Cail Corishev said...

But the basic talents that arty people possess have (as far as I've been able to tell in 35 years of hanging around various creative fields) almost nothing to do with IQ.

That's my impression too, as a high-IQ person who couldn't paint his way out of a cardboard box. High IQ helps you learn how to create art, just like it helps you learn anything else, but it doesn't make you good at it. When I started playing a musical instrument in 5th grade, I soon got moved up to play with the 7th-grade band because I learned to read music and memorized the fingering positions faster than anyone else. But it was all technical ability; I didn't play with any particular feeling, and I wouldn't know where to start in creating any new music. It was the same way in painting, sculpting, etc. I could quickly learn how paint colors mixed together in a mathematical sense, but had no feel for what colors would make a good sky.

I thought everyone saw artistic talent as a separate thing. Are there really people arguing that it's just a subset of IQ?

Cail Corishev said...

Shalizi's entire argument is really little more than a recap of Gould's: that a factor such as g rises pretty much inevitably in any phenomenon subjected to factor analysis, so why pretend that g is anything special or real?

To go back to his car analogy, if my car engine has twice the horsepower of anyone else's, I'm going to win most of the competitions I enter, whether they be drag races, all-day road rallies, small oval dirt tracks, tractor-pulls, or whatever. That much power advantage will overcome a lot of other disadvantages in each kind of race.

Now, if I understand correctly, this guy would say that's evidence that engine size doesn't matter, because it shows up too inevitably in every comparison?

I'm starting to think one of us is stupid, or thinks the other one is.

E. Gibbon said...

I just enjoyed math a lot more than memorizing names and dates in history class

Somebody simply had a bad instructor. Meaningful history is about the significance of events and the ways in which they develop over time (though you can't understand patterns if you don't know the basic facts).

C. Julius Caesar said...

i refuse to believe anybody on earth is actually that stupid. yet, there he is.

I answered this one long ago: quod fere libenter homines id quod volunt credunt ("generally speaking, people are happy to believe what they want to").

The t-factor (testosterone) trumps the g-factor every time!

Anonymous said...

Ray Sawhill- would you see artists as animals? Stronger than the rest of us, or more sensitive (wet nose, Jacobsen's Organ, hearing, just being closer to the ground so they pick up more).

jody said...

"Ray Sawhill said..."

it depends on which field of art it is. some are the way you describe. not very related to g and also narrow in scope. great ability in one art form does not translate to high ability in several other art forms. other fields of art are clearly g related, which leads to the dreaded "white guys stacking up at the top of the industry and monopolizing it" effect which our new cultural marxist leaders hate.

acting talent may be more widespread among humans and not highly g related, but writing, directing, cinematography, and effects talent are, and white men utterly monopolize those, and that's where a movie is made. not by the actors, who on their own, make nothing. passable acting ability is far, far more common than than the other abilities. actors come and go, there's an endless supply of them.

it's the same in the world of writing fiction and non-fiction books, and in the world of making video games. one group dominates those fields, and they are g related art forms. here, you just plain cannot participate in the art form without being smarter than average. the barrier to entry is a certain level of g.

even in music, some genres of sound are not very g related, but others are. writing a symphony, most forms of heavy metal, and trance techno, seem to the be sole province of guys with some above average g. you don't have to be a physics major but it seems like if you don't have a little on the ball upstairs you can't make good new original works in those specific genres. we have decades of evidence on that. in techno the various styles clearly break down by how smart the producer or DJs are on average. you don't have to be smart to make some decent house, jungle, or trap, but you do have to be smarter than average to make decent trance, trouse, or write techno pop singles for sale to american pop singers. dubstep probably falls in the middle, you don't have to be smart to make some decent dubstep, but being smart helps.

Alwaysright said...

It's a myth (promoted by people who are generally pretty high G themselves, and who like to think that the artists they study and promote are smarter than they often are) that art-talent is a cognitive talent

There are only two types of human abilities: physical and mental, so artistic talents would fall into one of those categories in factor analysis, with some zone of ambiguity for hybrid aptitudes like singing, dancing and drawing, which require both physical and cognitive ability.

The problem with artistic talent is it often reflects thousands of hours of practice rather than just innate ability and there is also a subjective component in judging the quality of art. It probably also reflects non-abilities like personality which affect creativity. Art is probably less g loaded than academic cognition, but it's still a cognitive ability.

Cail Corishev said...

Somebody simply had a bad instructor. Meaningful history is about the significance of events and the ways in which they develop over time (though you can't understand patterns if you don't know the basic facts).

Yeah, I discovered that once I got out of school and discovered how interesting history (and the various studies of humanity) really is. It took a lot of reading to make up for 13 years of memorizing boring crap long enough to pass the test and promptly forgetting it. I guess the silver lining is that I also missed some of the indoctrination that went along with it.

Anonymous said...

A guess as to part of what's happening here:

It stands to reason that multiple areas of the brain are recruited as part of cognition. This makes the "sampling model" intuitively appealing, while making g intuitively difficult to understand as a causal mechanism. However, the question of how cognition works and the question of what underlies individual differences in cognition are two quite separate questions.

The model which seems to fit the data presented here is that g ultimately reflects a collection of features of neuronal cell physiology as well as the physiology of higher-level parts of neuroanatomy that vary between individuals. Genetic effects on cell physiology and brain development tend to have brain-wide impacts, which get reflected in g. In contrast, one might imagine that various non-genetic effects would have more localized impacts on the brain and thus more variegated effects on variation in cognitive abilities. This causes the heritability of g to come close to 100%, while the heritability of composite IQ scores can be much less.

Unknown said...

The Nose Cone versus the tree
http://www.blogger.com/blogger.g?blogID=4624586630299165335#editor/target=post;postID=1530907615502361024

Ray Sawhill said...

Cail Corishev -- "I thought everyone saw artistic talent as a separate thing. Are there really people arguing that it's just a subset of IQ?"

There are a lot of people who hang out here at Steve's who see artistic talent as a subset of IQ.

Anonymous 11:01: "Would you see artists as animals? Stronger than the rest of us, or more sensitive..."

Where their creative/art-talents go, most artists are 'way more instinctive and sensual than they are intellectual, yeah.

Jody -- I agree that some IQ-style horsepower is certainly a help in some art-fields. Disagree with you about book-fiction and acting, though. Writers of book-fiction are often surprisingly non-intellectual people, and often just plain not very smart. (I remember attending a PEN conference decades ago and being amazed by how not-very-bright a lot of the worldclass authors were.) Creating book-fiction isn't much like writing extended nonfiction, which does tend to take a decent amount of brainiac-style horsepower. It's more akin to singing, dreaming, and acting than it is to anything intellectual. And, FWIW, I've got the highest regard for actors and performers. They're often seen as the icing on the cake, but in my experience they're the cake itself -- often the most creative people around. And (again, FWIW) the performing impulse often strikes me not as the final step in bringing something to life but as the primary impulse behind most art-creation.

Alwaysright -- "There are only two types of human abilities: physical and mental, so artistic talents would fall into one of those categories ..."

That's quite an assertion, that there are only two types of human abilities. Where do you get that? Doesn't correspond to my experience at all. In my experience, the artistic dimension is its own thing, as valid and as obvious a category as IQ-style cognitive horsepower and athletic talent.

Unknown said...

http://www.blogger.com/blogger.g?blogID=4624586630299165335#editor/target=post;postID=1530907615502361024

Anonymous said...

There are tradeoffs between some specific abilities (e.g., verbal and math). Such abilities may correlate negatively with each other (contradicting positive manifold), but only after removing general variance (g) from tests.

http://www.sciencedirect.com/science/article/pii/S0160289612001444 (Non-g residuals…)

The validity of g has been established for over 100 years. Higher g scores (extracted from diverse tests), and IQ scores (which are highly g loaded), are correlated with larger working memory, faster processing speed, reduced criminality, faster nerve conduction, better grades, increased longevity, better health … and so on.

Moreover, a test’s g loading (correlation with g) is positively correlated with its validity: The higher a test’s g loading, the better it predicts the aforementioned variables. The difference between forward and backward digit span is instructive: Both tests involve remembering digits, but backward DS requires that digits be recalled in reverse order (and therefore involves manipulating, as well as holding, information in memory): Guess which one has a higher g loading and a higher predictive validity?

Finally, g appears to be invariant across different test batteries (Johnson’s “Just 1 g” and “Still Just 1 g” articles), and across different factor analytic methods (Jensen & Weng’s “What is a good g” articles), as long as such methods allow for the extraction of variance common to the tests (e.g., don’t use varimax rotation).

The idea that g is an artifact would suggest a conspiracy on grand scale – involving, as it were, collusion among psychometricians, cognitive psychologists, neuroscientists, epidemiologists, geneticists, educational psychologists, test developers, and so on.

Rather than alleging such a conspiracy, a more useful objective would be to try to understand the causes and correlates of g. Fortunately, such a program of research is very much alive and thriving, and has been for more than 100 years.

John said...

It's not that g doesn't exist, it's just that g is not as important as the specific factors when it comes to intellectual ability.

g is simply the observation that someone who is good at one intellectual task tends to be better than average on others as well. That's either true or not, so that part of the controversy is easily solved. The question is what this means. And it doesn't mean much. Someone can be exceptionally good at math and just slightly above average in verbal skills. Such a person provides evidence for the existence of g, but we can easily see that his innate aptitude for certain intellectual tasks vary enormously. Thus the importance of g is seen to be trivial.

In other words, the tendency for convergence represented by g still leaves so much room for the non-g factors as to make the importance of g very small indeed.

The average person thinks of g as just "smarts" - in other words someone good at math can be just as good at verbal skills. He's got "smarts". But we see that that is not the case at all.

Then intellectual ability is a rather mysterious thing and can't yet be measured comprehensively by tests. Two people can have the same measured g and the same measure on the specific factor on any aptitude, put in the same amount of work, and yet perform wildly differently - we see this in the case of countries easily enough - showing that we are fools if we think our little contrived tests manage to capture everything about human intelligence, or even much about human intelligence.

It's a little known fact that before IQ tests were outlawed in the workplace, employers were very unhappy with its poor ability to predict performance and were casting about for alternatives.

Intelligence undoubtedly exists, but too many real world facts make clear that our IQ tests are only indifferent measures of it.

Cail Corishev said...

FWIW, I've got the highest regard for actors and performers. They're often seen as the icing on the cake, but in my experience they're the cake itself -- often the most creative people around.

Same here. I would submit for consideration all the actors, especially in TV, who eventually try their hands at writing and/or directing episodes and do a fine job of it. Ben Browder, Zach Braff, and Alan Alda come to the top of my head from different eras and genres, but there are many examples. Some, such as Jonathan Frakes, may end up more successful as directors than they were as actors.

That makes it seem almost as if anyone with a bit of ambition and attitude, whose acting talents get him into a position of influence, could be a decent director. How often does someone go the other way, starting out on the "creative" side and then discovering that he can act too?

Jody made a good point that some creative arts would be more IQ-based than others. Writing ought to be, since good writing demands a solid command of language, including grammar, vocabulary, paragraph structure, and so on. You can pick those up from reading lots of books, but again, high IQ will help you pick them up faster and better. I would guess that most of my favorite science fiction and fantasy authors are quite intelligent. Certainly Asimov, Douglas Adams, Terry Pratchett, Stephen Donaldson, Stephen King, and Zelazny are (or were) all high-IQ guys. Maybe contemporary navel-gazing fiction about (as Cartman would say) gay cowboys eating pudding could be written by average brains, but Pratchett makes puns about quantum physics, for cripes sake.

My guess would be that there's a cognitive ability we could call "creativity" which is separate from g-type ability. I don't think they necessarily contradict each other, the way people think geniuses can't have common sense, but I don't know that they correlate either. But when someone gets both creativity and high IQ, that's when magic happens.

Anonymous said...

As far as IQ vs artistic ability goes, here's my data point: I graduated as one of the top students of my high school. I also was at the top of my art class, but I didn't see as much future in that compared with math/science related learning, so that talent was never highly developed.

I also had a lot of talent musically (sufficient to excel early on), but did not have time to devote the hours and hours to get anywhere with it in my teens.

As far as the other kids who did well musically, they were without exception of above average intelligence, most were at or near the top of their class. But from what I could see, most did not really plan to go anywhere professionally with it.

I think there is a good chance that the smartest kids in school may have been capable of making better art than the drop-out types who generally pursued art, however, they also had the smarts to realise that a more secure and higher paid job was a better career choice.

Perhaps there is also the inclination to take into account. Hardcore, competitive maths types who see everything quantitatively, including games, will approach role playing games and such in that light. They will completely ignore the artistic side (e.g. this sword looks cool and has a cool background story!) in favor of the mathematical optimization (e.g. this sword deals the most damage per second, and also confers this and that ability etc.). Thus they don't get much practice in artistic thinking.

Gary C. said...

"acting talent may be more widespread among humans and not highly g related, but writing, directing, cinematography, and effects talent are, and white men utterly monopolize those"

Asians make films of superior innovation and quality in my reckoning. Only a handful of Hollywood filmmakers and no European talent can compare to the best that South Korea, Japan, Hong Kong and China have to offer.

Anonymous said...

Bickering about the existence of The Grail again. Thanks for putting it in the headline, Steve, makes 'em easier to skip past.

Anonymous said...

I read an article in Hustler once that said that the G Spot was a myth!

Svigor said...

Somebody simply had a bad instructor. Meaningful history is about the significance of events and the ways in which they develop over time (though you can't understand patterns if you don't know the basic facts).

That was my first thought, too. Either that, or a perceptual bias. History is the greatest story ever told. Boring, it ain't. And the names and dates are the least important and interesting part.

Perhaps it's relevant that history is my favorite subject, and math was always my least favorite, until I realized that math is the only truly challenging subject, the only one that forces you to use your brain like a muscle. Or me to use mine, anyway.

Svigor said...

Gary, yours is very much the minority opinion. It is a vocal minority, though. E.g., people who think The Raid was better than Dredd, even dramatically so. It is to laff. With no insult intended, I point out that these same people usually think the latter's plot is a rip-off of the former, even though the latter finished shooting before the former even started.

Svigor said...

I think "innovation" might be the key for those who favor Asian cinema. Like Japanese video games, "animation" (the proper term is "limited animation," really), and manga, Asian cinema is full of "innovation." I find the vast majority of these "innovations" to be crap. Others seem to find them interesting. To each his own.

David said...

Gore Vidal once said that Andy Warhol was the only genius he knew who had an IQ of 60.

"Art" is in the eye of the beholder. We make a category error if we swallow as "art" whatever art dealers and hip-hop promoters are pushing for the nonce. (You have your own fave examples of anti-art, I'm sure.) Defining "artistic ability" is a more treacherous business without that reservation.

I mean, Karen Finley and Michelangelo are both denominated as "artists," but how useful a definition of "artistic personality" could you glean from their commonalities?

Your neighbor who plays with clay may not be an artist. The kid who bangs a drum in his garage until he's 18 (or 38) may not be an artist.

Anonymous said...

Both tests involve remembering digits, but backward DS requires that digits be recalled in reverse order (and therefore involves manipulating, as well as holding, information in memory): Guess which one has a higher g loading and a higher predictive validity?

Generally tests which you would predict to overlap (conceptually) with many of the narrower than g factors tend to be high g loaded - its obviously less probable for someone to succeed at a g loaded test by skills involving a number of subfactors than it is by g, because the subfactors are uncorrelated or negatively correlated (after all, if they were not, they'd just join up in the factor analysis to become part of g).

If you are great at mental rotation, might just be because your mental rotation factor ability is pretty great, or it might be because you have a high g.

Who knows right?

But if you have to rotate blocks while doing so in a way that matches a particular verbal code, that involves a strong element or memorization, etc. then it becomes unlikely that you just happen to be good at all these things (high levels of a bunch of relevant factors, but not the ones that the test doesn't rely on) rather than having a high g. Thus high "g loading".

This is why there is the finding that complex tasks tend to be more g loaded!

Svigor said...

Bickering about the existence of The Grail again. Thanks for putting it in the headline, Steve, makes 'em easier to skip past.

Bickering about whether the Earth is round again.

FTFY.

John said...

"But if you have to rotate blocks while doing so in a way that matches a particular verbal code, that involves a strong element or memorization, etc. then it becomes unlikely that you just happen to be good at all these things (high levels of a bunch of relevant factors, but not the ones that the test doesn't rely on) rather than having a high g. Thus high "g loading"."

This statement well illustrates the tautalogical (no new information is offered) and hence trivial nature of g.

Since g is simple the observation the if you are good at one thing you will likely be good at another, then by definition a test that measures multiple things will be a better measure of g.

So you test a guy on a complex test (multiple abilities) and he does well; we say he has high g; translation into regular English - he does well on multiple mental tasks.

You test a guy on one task and he does well; we say could be g could be special factor; translation into regular English - we know he is good at this thing, but we don't know if he's good at other things as well.

All this is utterly, utterly trivial. If a heavily g-loaded test is one that tests for multiple mental abilities, then it says nothing new to call a high score on this test evidence of high g - the same idea can be more easily and clearly stated simply as....this guy does well on a bunch of mental measures, not just one.

Somehow, using the word g wraps it all in a mantle of scientific authority that makes it seem so much more significant.

The reality remains that some guys are good at all mental abilities, other guys have enormous variance in their abilities, but that there is a general tendency for people who are good at one mental task to be at least better than average on others but the differences are often enormous, rendering such an observation all but trivial, and in many cases this relationship does not even obtain at all.

In other words, g exists but is utterly trivial.

Anonymous said...

All this is utterly, utterly trivial. If a heavily g-loaded test is one that tests for multiple mental abilities, then it says nothing new to call a high score on this test evidence of high g - the same idea can be more easily and clearly stated simply as....this guy does well on a bunch of mental measures, not just one.
In a sense that's true. The tenets of factor analysis give you this one big variable, plus lots of smaller uncorrelated ones, rather than just a bunch of smaller correlated variables because that is how factor analysis works.

To be honest, in practice for most uses of IQ tests, this doesn't really make much difference.

The idea of g has really only two qualities about which I have reservations about:

- The search for a neurological g. There is some tendency for people to look for one neurological region in the brain where volume, activity, density, etc. is strongly correlated with g. This seems totally unlikely to me, as processing happens all across the brain, with modular regional specialisations. Regions that map to g might, at best, be regions that act as cross coordinators, but this is not likely to explain much of g.

- People tend to assume that g is "real" while non-g abilities are not "real". I don't know that mathematicians for instance (specialists in math) actually really do get more (or much more) of their professional success from having a high g than having high relevant cognitive subfactors. As you seem to suggest. But there is this assumption that g represents "real" cognitive ability while subfactors do not (that they reflect "simple primary sensing abilities" or something).

The g-loadedness of a test seems to me to more reflect how closely a task maps to an entire IQ battery in minature (although this is not strictly true, as IQ tests may be more imbalanced in terms of subfactor skills than particular tasks and thus show a lower g and higher subfactors than specific tasks). E.g. Raven's which involves memory, and both spatial and verbal decoding strategies, tends to come out as high g, because it recruits a mix of abilities. g-loadedness does not actually necessarily how difficult or important it is (because totally trivial and pointless tasks like Ravens can be "high g").

A lot of g's popularity comes from two factors

1) its high heritabilities - which is again not surprising, as we'd expect a factor which involves large swathes of the brain to have less random variance (and IQ is essentially the heritable plus the random) than factors which recruit small regions of the brain (law of large numbers again - across a wider range, random variance cancels out better than across a small range).

2) the idea that by finding and improving g, we don't have to make tradeoffs about what skills we regard as important. higher g is just better all around.

isomorphismes said...

But why are there only positive correlations among cognitive skills? That's hardly a trivial question.

Let's say questions meant to measure "verbal intelligence" and "quantitative intelligence" correlate with each other. This doesn't mean that one single latent factor underlies the success in each. It doesn't even mean that "verbal intelligence" or "quantitative intelligence" actually exist.

Rather than heeding the usual warnings about correlation/causation, assuming a latent unidimensional explanator does even worse than what those warnings are warning against.


Many things in life come with tradeoffs, such as risk v. reward.

How does that relate to the surrounding?

isomorphismes said...

Jim, your Monte Carlo test seems reasonable in the other direction (if random does as good as the model, the model's worthless). But just because this weak bar ("beats a monkey") has been met, how can this be seen as a strong indication that your model is the real truth?


(I realise that "monkey" is sometimes taken to mean uniform distribution, I'm just redefining monkey at a dart board to be some distribution with the desired characteristics)