g vs Modularity

Half Sigma has a great post describing attempts to determine g through testing and why the Flynn Effect is probably not measuring increases in intelligence.

One question that comes to mind is -

Doesn't the existence of g oppose some of the ideas of evolutionary psychology?

One of the presumptions of ev-psych is that rather than being a general purpose learning machine, the mind has specific 'modules' to handle specific tasks, such as learning languages or indentifying people who try to rip us off or finding the best mate possible. Having a general factor of intelligence seems to run counter to that notion.

Perhaps the mind is running different software programs (modules) but is still dependent on a central CPU (g)?

Share this

I determined g through

I determined g through testing. It's 9.81 m/s^2 but my cellphone will never work again.

Anyway, g assumes there is correlation between those different factors, I don't see how that contradicts the idea of separate modules.

A correlation is measured statistically on a population, it doesn't imply causation though. You mention the possiblity that the different modules are dependent on a common cause, that would be a possible explanation for the first principal component but it doesn't have to be.

Take fishes, there are big fishes, wide fishes, thick fishes... you have all kind of shapes, but whole in whole, if you measure every wish in all directions and look at the first principal component of your set, you'll find a factor that roughly gives you an intuition of the "size" of the fish. Why is that ? Because it doesn't make sense for a fish to extremely long and thin, or extremely flat (murenes and mantas exist but they're the exception not the rule)... it's not a wise shape for a given volume.

Similarly, the different modules of the mind may be completely independent, but people with extreme skills in a specific module spend too much energy on that part of the brain for a gain that's not really worth it because it's too specific. I see this a possible explanation why different skills might be correlated.




Makes sense

Makes sense

Not sure that works

Similarly, the different modules of the mind may be completely
independent, but people with extreme skills in a specific module spend
too much energy on that part of the brain for a gain that's not really
worth it because it's too specific. I see this a possible explanation
why different skills might be correlated.

I'm not sure that your argument from analogy works, because your different-body-plan fish are (presumably) fish from different species. Your analogy is that different species of fish tend not to depart too greatly from a certain stereotypical body plan, because a body plan that departs too much does not work very well. So your explanation is teleological (from fitness to actuality). The causal link from fitness to actuality is established through natural selection. That is, individuals which depart too greatly are culled, so that each species tends to remain close to the stereotypical body plan.

But we're talking about variations among members of a single species (humans). Supposing that two modules are, indeed, completely independent, genetically as well as mentally. Then each module has its corresponding gene, and since the modules vary in strength, the genes must therefore come in multiple alleles. To a first approximation, because sex mixes and matches the alleles of the parents, any given human being might get any combination of alleles. That is, an individual human might be strong in speech and weak in sports, strong in sports and weak in speech, strong in both, or weak in both. Sex (eventually - not necessarily in a given generation, but after a while) separates distinct genes from each other when the DNA of mother and father are remixed. Given the alleles that exist in the gene pool of the human species, then the combination of alleles that a given individual is born with is a product of a lottery called love. So the variation is not, at least not to a first approximation, quite the same thing as the variation we see between different species of fish. The reproductive isolation of distinct species protects the inter-species correlation between distinct genes which has arisen through natural selection; sex within a single species is continually destroying any intra-species correlation between distinct genes by remixing the genes. This correlation can of course be preserved if (say) human sub-populations form subspecies or races which are strongly isolated from each other. Races can serve the same correlation-preserving function that species serve in fish. We can talk about races if you like but for now I am assuming no races.

In case you don't understand why this matters, I'll try to explain. Suppose that there are two species of fish. One species is 1 inch long and 1/8 inch wide. The other species is 1 foot long and 1/8 foot wide. As you've explained, there is a reason that the proportions are similar. So far, no problem.

But now imagine that there is a single species of fish such that some individuals are 1 inch long and other individuals are 1 foot long, and some individuals are 1/8 inch wide and other individuals are 1/8 foot wide. Suppose furthermore that these properties are totally independent, with separate genes encoding each one independently. Then the next generation will be a mix-and-match of the length and width properties. Some individuals will be 1/8 inch wide and 1 foot long, and others will be 1/8 foot wide and 1 inch long.

Now, to be sure, the individuals with really bad proportions will die quickly. But importantly, before they die, they will be born. So they will exist, at least for a time. And furthermore, their dying will not prevent the next generation from producing a similar mix of well-proportioned and really-badly-proportined individuals. The species will continue to produce this mix of good and bad until either it somehow manages to split into two separate species (a small one and a large one), or one of the two sets of alleles (either the large set or the small set) is eliminated from the gene pool of the species.

Now, among humans, I do not notice any massive die-off among born humans. So I think it is safe to presume that the alleles in the human gene pool are, for the most part, universally compatible. Hopefully we can avoid complications such as the problem of dominant and recessive traits and the problem of incest. Complications aside, the alleles that actually survive in our gene pool can be mixed and matched and produce a viable human.

Thus, if there are different alleles for sports skills and speech skills, then they are all pretty compatible. Call the alleles for sports skills T and the alleles for speech skills C. Call them, specifically, T1 and T2 (two alleles for sports skills) and C1 and C2. Then evidently people with all of the following combinations are reasonably fit: T1C1, T2C1, T1C2, T2C2. This is, mind you, all possible combinations. So, among humans, if there are separate modules, and if the modules are truly independent, and if the quality of those modules varies among humans, then any given human can have any combination of those module qualities and still be a reasonably fit human being. But aside from that, because of sex, even if (say) T2C1 is unfit, then even so some newborns should have T2C1 even if they die quickly, because the newborns should display the full gamut of mathematically possible combinations of module-alleles, with a frequency derived from the frequency of the individual module-alleles. (You might dispute the conclusion by arguing that there is some sort of dying-off of unfit babies at conception or in the womb, or alternatively, you might talk about subpopulations - races or subspecies of humans - which complicate the picture - if you want to introduce such things we can of course talk about how they affect the conclusion.)

So, in short, there will not be some correlation, within the human species, of modules. We will not see some humans with all-small-modules, other humans with all-large-modules, and few humans in between. We should see, instead, humans with pretty much any combination of small and large modules - if the modules are truly independent.

In contrast, when we compare distinct species of fish, because of the genetic isolation of different species from each other, then we will see correlations among even the genetically independent features. Alleles that are incompatible with the remainder of the genome will be removed from the gene pool of each species.

The fact that the analogy

The fact that the analogy does not work intraspecie crossed my mind, but I thought it was ok, and then I was not sure, then I thought it definitely worked and now I'm not sure. Anyway.

Assume you have two independent functions, A and B, which you can develop, A is useless without B and B is useless without A, but both have a cost, and carrying an extra skill in A that is not matched by a skill in B is terribly bad. 

What will happen is that the genes for A a and B will become a single gene, as anytime it is broken up, it severely handicaps its bearer. Maybe they'll move closer to each other. In fact, a genre C with an allele that comes in to prevent B from expressing itself if it is not matched by A and vice versa would be very successful.

In the population you will observe different levels of skills, but the A skill will always match the B skill. You might argue that they are not independent because of C, but the key is that C may have nothing to do with a causal  factor like "processing power", it is just here to prevent a disequilibrium in skills.


Okay, so they might not be genetically independent (share a single gene, I'll call it a kitchen sink gene ("being or made up of a hodgepodge of disparate elements or ingredients")) while still being mentally independent. This strikes me as unlikely, but as far as I know possible. Unlikely because we're talking about evolutionary psychology, which has come up with a rather detailed list of mental contents (rather too detailed, if you ask me), and even if all that stuff is separately encoded, can it really all fit in a strand of DNA short enough not to be frequently divided by sexual reproduction? And what's the likelihood that the human species actually has separate kitchen sink alleles, one allele for one large complex of distinct mental modules, and another for another large complex of distinct mental modules? Furthermore such a kitchen-sink gene seems out of the ordinary when compared to the other genes I've heard of. Furthermore, if there really is a kitchen-sink gene with kitchen-sink separate alleles, we might expect something other than a one-to-one correspondence of parts between the alleles. That is, why would we expect one allele to have A, B, C, D, etc., and another allele to have A', B', C', etc., all just like A, B, C, etc. but slightly better? Why wouldn't it have a completely different set of stuff - G, H, I, J? After all, an individual human either expresses one allele or the other. What keeps the two alleles in lock step with each other like that? Is the pressure of a common environment really that strong? I'm sure it's strong but this seems excessive. It just seems to be an unlikely explanation. And is there really enough time in human evolutionary history for getting the genes for the different hypothetical modules all together in a nice little gift-wrapped package? If they work so poorly individually and work well only together, then an initial phase in which they were separated and then later migrated closer seems to be unsustainable by the very assumption that they don't do well unless they travel together.

My intuition is that such a thing is too brittle to be at all likely. Admittedly I'm not able to come up with a solid argument backing up the intuition.

One thought that occurs to me, but for which I have no proof, is that joining all those different module-genes into one gene might interfere with their ability to further evolve separately. That is, suppose that the modules-genes are A, B, C, and D, and A', B', C', and D' are better versions of these - better in the sense that eventually they will replace them, so I'm talking about evolution, not just coexistence of alleles like before. Suppose that there are a thousand people, and in ten people A mutates into A', and in ten people (probably not the same people) B mutates into B', and so on. Over time, A' will replace A, B' will replace B, etc., as the initial mutants spread. After several generations, 900 people will have A', 900 will have B', etc. So there will be many people who have A', B', C', and D'. Great. We have evolution!

But now suppose that there is a single gene, ABCD, which produces all those features. Suppose that, as before, some people have mutations: in ten people ABCD mutates into A'BCD, in ten people ABCD mutates into AB'CD, and so on. Over time, the mutations will (by hypothesis) replace the originals, but look at what the mutations are! They are A'BCD, AB'CD, and so on. Whoops. There's nobody with A'B'C'D'.

So, maybe that is a reason to think that genes will tend not to be kitchen-sink genes.

Module development doesn't need to be restricted in time

"And is there really enough time in human evolutionary history for getting the genes for the different hypothetical modules all together in a nice little gift-wrapped package?"

What makes you think the modules were developed solely in "human" evolutionary history? 

BTW, I think there is some confusion here with the types of modules that Steven Pinker is thinking of and this idea that there are sports, math, verbal, etc. type modules.    Evolutionary psychologist are trying to explain consciousness via modules not differential talents.

I can infer that my dog has a high degree of consciousness, at least how I define it, by the fact that it navigates her environment in a way that shows she has a complex model in her head.   For instance, she doesn't mistake my fingers for dog treats even if they smell of them.   Can't say the same thing about my pet snakes.   Yet even the snakes at a certain level seem to have a mental model of their world.

So drop it

What makes you think the modules were developed solely in "human" evolutionary history?

So remove "human". It's not key.


His scenario is more likely with recessives

Ok, now change Aurthurs scenario so that A and B are recessive genes and it's perfectly possible for his scenario to happen.   Deleterious recessives can be maintained  in the gene pool for long periods of time because they are not selected against until they are homozygous.    

I don't think his scenario is necessary to evolve a modular system which can be effected in lockstep by other genes.    Genes can have this type of effect merely as a side effect of some other adaptive feature being selected for.

Many genes are polymorphic and all that need happen is that there are genes that are being selected for other reasons that increase or decrease overall brain function.   Also, since we are a species that survives via division of labor we in a sense live in multiple environments in parallel.    Such conditions can support multilple alleles in the same population due to differential selection pressures depending on role.

Our muscle systems, skeletons, and our fat storage systems are modular, yet there tends in most people to be a correlation between strength, bone length, and fatness throughout the body.   Mesomorphs tend to be more muscled in all muscle units, tall people tend to have longer bones in general and not extra long necks, etc.

Besides we do have examples of people with genetic differences having vastly improved intelligence in one area and yet diminished capabilities in other areas, idiot savants.   If our intelligence were a general property of having more neurons and not due to modular brain structure then you wouldn't expect increased functionality in one area paired with diminished capacity in another like you do with idiot savants.   That is unambiguous evidence that the general capacity theory is false.  


Not just A and B

As I understand it, A and B are just for simplicity of explanation. The modules are numerous. Say, conservatively, 50 separate mental modules, each one represented by a different recessive gene. It just seems really, really unlikely 50 modules, all of which need every single one of the other 49 to be at all useful, to appear (either separately or together), and then gather together into a single kitchen sink gene that keeps them together. I'm not sure that making them recessive makes this any easier.

When I think of the evolution of something complex, I think of the eye, which has a clear path from absolute simplicity (maybe a patch of skin that is sensitive to light) to the complexity and delicacy of the eye. I see an evolutionary path here. I have a harder time making sense of the independent-but-interdependent-module hypothesis as something with an evolutionary path.

Our muscle systems, skeletons, and our fat storage systems are modular,
yet there tends in most people to be a correlation between strength,
bone length, and fatness throughout the body. Mesomorphs tend to
be more muscled in all muscle units, tall people tend to have longer
bones in general and not extra long necks, etc.

And the analog of my own favored hypothesis is that there is not an independent gene for each muscle of the body determining every aspect of it. In contrast, I am arguing against the hypothesis that every muscle in the body has its own independent gene, its own independent programmed strength that cannot be tweaked on a global level by a genetic change but must be independently tweaked, and that the hundreds of genes, each one for the hundreds of muscles, have gathered together into one kitchen sink gene. Can you imagine how difficult it must be for such a creature to evolve into a stronger or a weaker creature? Every independent muscle must evolve separately! It would be much easier for a creature to evolve if there were some global property that could be tweaked to change its general level of strength. This is what I think is the case, and it is the denial of this which I argue against.

In contrast, my guess is that there are many genes that control different aspects of the muscles, and that if you tweak a simple gene you will likely get a system-wide change in body type. Whereas I am arguing against the hypothesis according to which, if you tweak a single gene (one sitting inside a kitchen-sink gene, which is a tightly packed length of DNA with all the muscle genes packed together), then precisely one muscle in the body will change its strength, leaving the others unaffected.

This is what I am arguing against. I'm not at all sure what you think I'm arguing against.

Besides we do have examples of people with genetic differences having
vastly improved intelligence in one area and yet diminished
capabilities in other areas, idiot savants. If our intelligence were
a general property of having more neurons and not due to modular brain
structure then you wouldn't expect increased functionality in one
area paired with diminished capacity in another like you do with idiot

As I already argued, this is a false dichotomy. It is not either-or. It is not either nothing-but-general-intelligence or distinct-modules-with-no-general-intelligence.

To recap the argument in terms of muscles, my initial point was that if each muscle were independently encoded, then the fact that humans come in different body types (mesomorphs, etc.) would immediately result in mongrel body types in a single generation, with people have all sorts of different mix of strong and weak muscles. The response to that was to speculate that all muscle genes were gathered together in a single spot so that the genetic mixing that occurs with sexual reproduction would tend to keep those genes together. My main response to that was that this seemed unlikely as a product of evolution. I gave several weak objections and one stronger one (I thought). I thought it would tend to interfere with further evolution if all the genes were together. Projecting this backwards in time, this would have interefered with evolution. That is, I think the proposed kitchen sink gene would tend to be stuck in time and be bypassed by evolution.

That is unambiguous evidence that the general capacity theory is false.

Straw man. Who is arguing for a general capacity theory? You, yourself, stated succinctly my own view on the relationship between modules and general intelligence:

I don't see why. Computers are modular and yet a "general factor" like transistor speed boosts their performance.

What I was arguing against was the complete absence of any such thing as a general factor. That was the idea that was being set forth. And while, in principle, the idea makes sense (and should be pointed out and I'm glad it was pointed out), I've been arguing that it's a bit more difficult to implement in evolutionary practice.



For one thing, we need goals, and intelligence isn't itself a goal, it's a tool to get to goals. Without goals, we'll just sit and do nothing with our intelligence, no matter how intelligent we become. We might have general intelligence, hooked to a reward/punishment system that systematically pushes us to pursue certain goals and not others. The tiny bit of evolutionary psychology that I've read seems focused on our different goals. Evolution might tinker with our reward/punishment system, adjusting what things make us feel good or bad. Examples of aspects of human psychology that evolutionary psychology has attempted to illuminate: why we like sugar and fats so much when they're so unhealthy (answer: they're rare in the ancestral environment), why do women prefer X, Y, and Z in men and why do men prefer A, B, and Z in women, why are men competitive, why are women risk-averse - and so on and so forth, most of them concerning preference, which is arguably distinct from intelligence.

Since training develops mental muscles, then choice of what to focus on can change what mental muscles are most developed, and that choice can in turn be affected by preference. Someone who loves music will focus on music and become a better musician than someone who is indifferent to music. Since much of mental development happens in early childhood, these differences in focus might easily escape the notice of researchers. We might observe that some young child simply seems particularly "innately" adept at something, when in fact, unbeknownst to us, he has been obsessed with it and become expert with it during his young and mostly invisible life.

The plasticity of the brain of a young person (if a part is damaged then other parts can sometimes replace its function) underlines the centrality of learning to brain function.

Since intelligence probably rests on a set of different factors that cooperate, then by adjusting the factors, intelligence can be tweaked to optimize it for different tasks. Take a shoe and make it higher, and it becomes a boot, losing flexibility but gaining protection - a trade-off. And, analogously, tweaking the brain might make it better suited for certain tasks while simultaneously worse-suited for other tasks. This idea is not the same as the idea of separate, specialized modules.

And it need not be either-or. It need not be either general intelligence or a set of modules. It might be a bit of both. General intelligence might be straightforwardly enhanceable by increasing brain size, and from the little I've read, it actually is enhanceable that way, which suggests that the brain knows what to do with any extra brain cells that are thrown at it. But in addition to this, there might be particular tricks which the brain of, say, a rat simply will never acquire no matter how much you boost its brain size. It may require the addition of topologically distinct features in order to get a rat to learn language (say).

Not really counter

"Having a general factor of intelligence seems to run counter to that notion. "

I don't see why.   Computers are modular and yet a "general factor" like transistor speed boosts their performance.

The intuitive way to think

The intuitive way to think about it is that g represents maximum potential capability at any given task. This squares away with the dual observations everyone has probably made at some point: that while having a low IQ seems to doom a person, having a high IQ is no guarantor of achieving anything of note.

Jonathan, have you read

Jonathan, have you read Malcolm Gladwell's recent New Yorker essay on Flynn and IQ? He makes a very similar point:

“If the everyday world is your cognitive home, it is not natural to detach abstractions and logic and the hypothetical from their concrete referents,” Flynn writes. Our great-grandparents may have been perfectly intelligent. But they would have done poorly on I.Q. tests because they did not participate in the twentieth century’s great cognitive revolution, in which we learned to sort experience according to a new set of abstract categories. In Flynn’s phrase, we have now had to put on “scientific spectacles,” which enable us to make sense of the WISC questions about similarities. To say that Dutch I.Q. scores rose substantially between 1952 and 1982 was another way of saying that the Netherlands in 1982 was, in at least certain respects, much more cognitively demanding than the Netherlands in 1952. An I.Q., in other words, measures not so much how smart we are as how modern we are. 

I've been meaning to, but I haven't yet

I'd disagree slightly with that last sentence:

An I.Q., in other words, measures not so much how smart we are as how modern we are.

I think the rise in IQ measures how modern we are. There's still a large component that measures how smart we are.

Right, I think that's what

Right, I think that's what Gladwell is getting at in that excerpt, so he would agree with your quibble.

More Race Science

G is a surrogate for personal worth. It is a vital element in social status. Of course everyone doesn’t have their G score tattooed on their forehead. Thanks to psychometric testing studies all you have to do is look at the color of someone’s skin.
I have devised different tests that measured different characteristics that still show the same thing. I have done the research myself and it confirms what we all knew.
Alligator wrestling in Louisiana was studied. Near New Orleans sociologists arrived at the bayou to administer the tests. It was determined that a white Cajun can on average whip an alligator that weighs 50 lb more than an alligator wrestled by an African American. It has also been determined that Cajuns can beat the average Black person at pie throwing.
Once again when Blacks are put to a test designed by White people, guess who wins? Is it hereditary or environmental? In the Minnesota Study similar results were obtained. Whites outperformed Blacks in log rolling and in chopping down trees. According to the Flynn effect each generation has improved on these tests but Blacks still lag behind Whites. Heredity must play some part.
Scientists have determined that the Cajuns have more slow twitch muscle fibers and an average two length shorter humerus than Blacks. This gives the Cajuns more leverage and a greater ability to wear down the alligator in a fight. It also appears to make them better pie throwers and lumber jacks.
Things work just the opposite in the city where Blacks surpass Cajuns and other Whites in basketball and boxing.

Let's do the following, we

Let's do the following, we blindly pick a set of random children. Then we agree on a wide array of "success" for criterion in later life... it can be anything, but what matters is really that they are many many criterions... make your choice... we can take wage at age 35, highest level of education achieved, criminality, couple stability, life satisfaction... I don't want to impose you a specific set, go and pick a set that you feel provides an honest, common sense  description of a "successful life". Then, we make bets on each of these children.

I get to look at their g factor. You get to look at a test to help you in negociating the bets, what do you choose to look at ?

Reply to Arthur B

Parenthetically, the Minnesotans and Cajuns may be related in another way. I believe many lumberjacks in Northern Minnesota were of French/ Canadian ancestry. More proof of my hypothesis. Also note the absence of French basketball players and boxers. It seems like an air tight case that there is an undefined factor G that acts as a differentiator between French immigrants and African Americans.


Someone may object G is just a second order derivative of cultural or economic differences. The primary factors I picked in my example were just not as emotionally loaded as the ones you picked such as “ wage at age 35, highest level of education achieved, criminality, couple stability, life satisfaction.. “


In the US there was a long process of acculturation that immigrants such as the Irish, Italian and Polish went through. All of them, especially the Irish and Italians were thought to be lazy, violent, immoral louts with a tendency to alcoholism, gangsterism and corruption. I think there G score would be lower than native “ White people”, though they didn’t test anyone then. Now Poles ,Irish and Italians are “White people.”


Perhaps G is just an indicator of acculturation to upper and middle class Western ways. This, of course doesn’t exclude other explanations, but historically genetic and other essentialist explanations have been heavily tainted by racism and ethnocentrism, so we must be very careful here



Uh ? How exactly is this a

Uh ? How exactly is this a reply to my challenge ? 

To make your experiment

To make your experiment truly "scienfitic", let's do the following: Posit an alternate parallel universe identical to ours in all respects but one: the skin pigment of white people and black people, since the beginning of time, is reversed. History, culture, and social views remain the same as ours, with pigment modified.

Now, we blindly pick a set of random children from our "normal" universe, the one you and I currently inhabit. We transport these children to the posited alternate universe described above. Then we agree on a wide array of "success" for criterion in later life... Then, we make bets on each of these children...