IQ and general intelligence

[Luke Skywalker]

Lately I've gotten really interested in psychometric research, particularly that pertaining to intelligence. I think it's fascinating that a single construct seems to have so much explanatory power, and that so much sociological research just pretends it doesn't exist.

Essentially, (and there may be professionals here that can correct me/etc.) if you give people a battery of different cognitive tests, you'll find that their performances on each are correlated, .ie, people who do well on one type of cognitive task also do well on others. Through some statistics you can extract a common g factor that encodes this positive manifold. This general factor is what IQ tests are trying to measure; better tests are the ones that are more g-loaded, .ie correlated with g. It turns out that the g-loading depends on how complex the task is, but not on what particular academic subject or activity it comes from.

There's some speculation that this factor is just a statistical artifact and could just be the product of tasks drawing from an overlapping pool of independent abilities, but that wouldn't explain why the g-loading is so indifferent to the type of test, why it correlates with parts of brain physiology, elementary functions like reaction times, etc. Generally speaking, if you're "smart", you're going to be smart at a wide variety of things; there are exceptions, of course, like people who are dyslexic and have specific learning disabilities, but they don't refute the general correlation. Likewise, it seems as though practicing at a particular task may make you better at that task, but doesn't have far transfer into entirely unrelated fields, .ie, you can get better at skills but not at g, given adequate nutrition, etc.

And despite conventional wisdom, IQ/g not only predicts things like academic performance, it also predicts a variety of life outcomes, including job performance (better than conscientiousness or work experience), health, relationship success, etc. better on a whole than any other variable. (the correlation is most apparent when you correct for range restrictions) Additionally, it seems to be a rather stable construct by adulthood, and your IQ eventually correlates with your biological parents better than it does with your adoptive ones.

The Gladwell idea is that this matters up to a certain point, but beyond a certain IQ you don't really get much more. At least for academic and most professional attainments, this simply isn't true.

It's interesting because it's so politically incorrect to put value on human intelligence, given its elitist implications and some evidence of average differences between ethnic populations and genders. But the evidence - and I have read the criticisms - seems to be overwhelming, and well accepted by psychometricians. The problem I have, then, is how many people try to use disparate outcomes in various endeavors to be automatic indicators of discrimination/ineffectiveness/etc., and don't even consider the possibility of genetic factors at play here. To be clear, I'm not euphemistically referring to race - a less sensitive example could be the commonly cited statistic that people with higher levels of education make more money, but nobody bothers to control for how smart/capable they already are (the correlation doesn't vanish as a whole but does become more nuanced). Likewise, a lot of people blame schools and parents for outcomes that can at least partially be traced to different levels of natural ability among students.

Thoughts?

[Simon_Jester]

Lately I've gotten really interested in psychometric research, particularly that pertaining to intelligence. I think it's fascinating that a single construct seems to have so much explanatory power, and that so much sociological research just pretends it doesn't exist.

Please be advised that the world is inevitably more complicated than it appears after one or two years of college. If you want to remain on the path from "know nothing" to "know a lot," keep looking out for complicating variables; don't try to impose a one-size-fits-all explanation on things. There are a LOT of such explanations, most of which have or had a small core of very devoted people who use it to explain everything... and they're mostly wrong.

Essentially, (and there may be professionals here that can correct me/etc.) if you give people a battery of different cognitive tests, you'll find that their performances on each are correlated, .ie, people who do well on one type of cognitive task also do well on others. Through some statistics you can extract a common g factor that encodes this positive manifold. This general factor is what IQ tests are trying to measure; better tests are the ones that are more g-loaded, .ie correlated with g. It turns out that the g-loading depends on how complex the task is, but not on what particular academic subject or activity it comes from.

Stop and think for a moment about what this proves. It may well be in part a measure of innate biological qualities of your brain. But it may also be a measure of how much overall stimulation and development your brain has received.

I mean, you could construct an equivalent of g (let's call it 'p' for 'physique') to measure the fitness, health, and overall strength of human beings. And there would be a significant positive correlation- the strongest people tend to be above average in terms of stamina and other measures of fitness. We could, with sufficient statistical analysis, find a single measure for "how fit are you," and define good 'fitness tests' (such as, say, a decathlon) that measure overall 'p' coefficient rather well.

Sure, there'd be the equivalent of idiot-savants, such as people who can bench-press huge weights but are clinically obese, but we could ignore those outliers.

Would it be accurate to say that 'p' is inborn, and intrinsic, on the grounds that people who have high 'p' in childhood often have high 'p' throughout their lives? Because they do; remaining physically fit is habit-forming and like any other habit, exercise remains most persistent for people who have been doing it since their youth. But while clearly genetics plays a role in determining how strong and fit you are, only a complete moron would claim that ongoing exercise, good personal health habits, and so on don't play a major role.

There's some speculation that this factor is just a statistical artifact and could just be the product of tasks drawing from an overlapping pool of independent abilities, but that wouldn't explain why the g-loading is so indifferent to the type of test, why it correlates with parts of brain physiology, elementary functions like reaction times, etc. Generally speaking, if you're "smart", you're going to be smart at a wide variety of things; there are exceptions, of course, like people who are dyslexic and have specific learning disabilities, but they don't refute the general correlation. Likewise, it seems as though practicing at a particular task may make you better at that task, but doesn't have far transfer into entirely unrelated fields, .ie, you can get better at skills but not at g, given adequate nutrition, etc.

I question how well this applies to the sort of things people do in early childhood. Because it invites the question, how do you even measure the 'g' value of a twelve month old child? How do you know the child's true 'g' value hasn't changed between the ages of, oh, one year and five years? How do you normalize test scores so that scores can be compared meaningfully among children who possess such vastly different cognitive skills? Especially when one of them cannot speak meaningfully, let alone read or write, and the other can do all these things?

We know that children who are not adequately stimulated in childhood wind up massively dysfunctional. So much so that their 'g' value almost has to suffer, if nothing else because of behavioral issues that make it difficult to test them meaningfully.

I don't feel confident trying to rule out the reverse- that children who are highly stimulated in constructive ways develop higher 'g' coefficients than less competently raised twins would.

After all, psychometric tests measure things that are learned skills in their own right, such as pattern recognition, following step by step directions, and so on.

And despite conventional wisdom, IQ/g not only predicts things like academic performance, it also predicts a variety of life outcomes, including job performance (better than conscientiousness or work experience), health, relationship success, etc. better on a whole than any other variable. (the correlation is most apparent when you correct for range restrictions) Additionally, it seems to be a rather stable construct by adulthood, and your IQ eventually correlates with your biological parents better than it does with your adoptive ones.

Question for you:

Does this prove that nature is winning out over nurture, or does it prove that it's hard for adoptive parents to provide nurture that competes fully with that provided by biological parents who don't get their children taken away? Not impossible, necessarily, but hard?

It's interesting because it's so politically incorrect to put value on human intelligence, given its elitist implications and some evidence of average differences between ethnic populations and genders.

Honestly, no. The problem is that we really are not ready to use psychometrics to determine to what extent 'g' is due to biological innate factors, and to what extent it is due to quasi-heritable cultural factors.

When you're a member of an ethnic group where you are the first person in your family to go to college, it's hard to be sure whether you're first because your ancestors were poor, or whether they were stupid. Since poverty is an obvious explanation, and the null hypothesis is "NO difference between races in basic intelligence," rather than "lots of difference..." I suspect poverty plays a significant role.

Likewise, a lot of people blame schools and parents for outcomes that can at least partially be traced to different levels of natural ability among students.

What people justly blame schools and parents for is for failing to make the most of what's available, or for writing children off as 'uneducable' or 'hopeless.' Especially when with discipline and effort and a program oriented toward their talents they could make something useful of themselves. Maybe not in a field requiring great abstract thought and a commanding intellect, but somewhere in some way. If you don't give them what they need, though, they tend to turn into delinquents or useless semi-employable drones.

[aerius]

Summary: Congratulations, you've found a possible/likely correlation, now you get to prove a causation. Have fun.

[Luke Skywalker]

Please be advised that the world is inevitably more complicated than it appears after one or two years of college. If you want to remain on the path from "know nothing" to "know a lot," keep looking out for complicating variables; don't try to impose a one-size-fits-all explanation on things. There are a LOT of such explanations, most of which have or had a small core of very devoted people who use it to explain everything... and they're mostly wrong.

Sure, although I'm interested in this variable's ability to explain lots of sociological phenomenon statistically or even causally, which doesn't really relate to my life/anecdotal experience beyond having enough of a stat background. But you're generally right.

Stop and think for a moment about what this proves. It may well be in part a measure of innate biological qualities of your brain. But it may also be a measure of how much overall stimulation and development your brain has received.

Twin adoption studies show a greater correlation of your IQ with your biological than adoptive parents across socioeconomic classes.

I mean, you could construct an equivalent of g (let's call it 'p' for 'physique') to measure the fitness, health, and overall strength of human beings. And there would be a significant positive correlation- the strongest people tend to be above average in terms of stamina and other measures of fitness. We could, with sufficient statistical analysis, find a single measure for "how fit are you," and define good 'fitness tests' (such as, say, a decathlon) that measure overall 'p' coefficient rather well.

Sure, there'd be the equivalent of idiot-savants, such as people who can bench-press huge weights but are clinically obese, but we could ignore those outliers.

Would it be accurate to say that 'p' is inborn, and intrinsic, on the grounds that people who have high 'p' in childhood often have high 'p' throughout their lives? Because they do; remaining physically fit is habit-forming and like any other habit, exercise remains most persistent for people who have been doing it since their youth. But while clearly genetics plays a role in determining how strong and fit you are, only a complete moron would claim that ongoing exercise, good personal health habits, and so on don't play a major role.

You can easily do this and measure its heritability. For IQ, by late adulthood it tends to be .5 - .8, a minority from shared family environment, and then a minority from unshared environmental factors.

I question how well this applies to the sort of things people do in early childhood. Because it invites the question, how do you even measure the 'g' value of a twelve month old child? How do you know the child's true 'g' value hasn't changed between the ages of, oh, one year and five years? How do you normalize test scores so that scores can be compared meaningfully among children who possess such vastly different cognitive skills? Especially when one of them cannot speak meaningfully, let alone read or write, and the other can do all these things?

You can construct different tests and then see how well they correlate with later, better established ones. They've done this, although the younger the person, the noisier the results.

We know that children who are not adequately stimulated in childhood wind up massively dysfunctional. So much so that their 'g' value almost has to suffer, if nothing else because of behavioral issues that make it difficult to test them meaningfully.

I don't feel confident trying to rule out the reverse- that children who are highly stimulated in constructive ways develop higher 'g' coefficients than less competently raised twins would.

After all, psychometric tests measure things that are learned skills in their own right, such as pattern recognition, following step by step directions, and so on.

Neglect and malnutrition can certainly lower IQ (and other things). The reverse, really high quality stimulation, thus far doesn't seem to have a proportionally amazing benefit, at least on g - programs like Head Start may contribute to certain behavioral advantages, but its cognitive boosts fade away by late elementary school.

Question for you:

Does this prove that nature is winning out over nurture, or does it prove that it's hard for adoptive parents to provide nurture that competes fully with that provided by biological parents who don't get their children taken away? Not impossible, necessarily, but hard?

That's irrelevant to the statistic, which compares the extent to which their IQ scores correlate with biological vs. adoptive parents.

\Honestly, no. The problem is that we really are not ready to use psychometrics to determine to what extent 'g' is due to biological innate factors, and to what extent it is due to quasi-heritable cultural factors.

I suppose there is a range restriction on the environmental variable with respect to performing all these tests in first world countries, but that hardly precludes figuring out the genetic component of things, especially if we can independently correlate g with certain genetic code (which they have, and are trying to do more thoroughly).

When you're a member of an ethnic group where you are the first person in your family to go to college, it's hard to be sure whether you're first because your ancestors were poor, or whether they were stupid. Since poverty is an obvious explanation, and the null hypothesis is "NO difference between races in basic intelligence," rather than "lots of difference..." I suspect poverty plays a significant role.

Sure, but it's also possible that their poverty was partially caused by their, well, relative stupidity, at least in the generations following the civil rights act and a relatively universal education system. And your g correlates with your future job performance and IIRC your income better than your family's socio-economic status.

What people justly blame schools and parents for is for failing to make the most of what's available, or for writing children off as 'uneducable' or 'hopeless.' Especially when with discipline and effort and a program oriented toward their talents they could make something useful of themselves. Maybe not in a field requiring great abstract thought and a commanding intellect, but somewhere in some way. If you don't give them what they need, though, they tend to turn into delinquents or useless semi-employable drones.

Sure, but if you want to measure how well you're making do with what's available, you have to control for that, and most commentators aren't willing to.

[Luke Skywalker]

Summary: Congratulations, you've found a possible/likely correlation, now you get to prove a causation. Have fun.

"proving" causation is probably only possible via genetic linking (which has been done to some extent - given time I can pull sources), but it's not that difficult to produce statistics that make causation the most logical answer. For one thing, we can look at twin adoption studies, where the adoptee's adult IQ correlates more with that of their biological parents than their adoptive ones - this makes the possibility of socioeconomic and cultural factors being the primary cause of IQ (at least within the spread of 99% of a western democracy) incredibly unlikely. We can then observe the correlation of g with relevant life outcomes while controlling for every other measurable variable, and we find that there remains something substantial. At that point, causation is the most likely explanation.

[Flagg]

My IQ is in the upper 130's lower 140's depending on the test. Check out my life story.

[Luke Skywalker]

My IQ is in the upper 130's lower 140's depending on the test. Check out my life story.

I know that you're kidding, but just to jump off of it: I think aside from focusing on anecdotes, a lot of us might understate the importance of IQ because we honestly see a lot of other factors mattering more. This may be true in a statistical sense because we already self-select our peer group for people that fall within a certain intellectual range. We might say "Jack is so much smarter than Bob, but Bob gets better grades and ended up getting a better job, I guess hard work matters more", but both Jack and Bob might have already been pretty smart fellows, and the difference between them was actually small relative to the general population. If you take a sample of everyone, you might find a more drastic difference.

[Simon_Jester]

Twin adoption studies show a greater correlation of your IQ with your biological than adoptive parents across socioeconomic classes.

Does this control for the possibility that adoption itself is an issue in play here? Also, adoptions are not necessarily done at an age prior to the early childhood brain development, how young were the adoptees?

You can construct different tests and then see how well they correlate with later, better established ones. They've done this, although the younger the person, the noisier the results.

Thing is, it's precisely among the youngest children that

Neglect and malnutrition can certainly lower IQ (and other things). The reverse, really high quality stimulation, thus far doesn't seem to have a proportionally amazing benefit, at least on g - programs like Head Start may contribute to certain behavioral advantages, but its cognitive boosts fade away by late elementary school.

Behavioral advantages, especially if persisting over time, may well have consequences for overall life outcomes that are indistinguishable from an increase in 'g.' A smart person with severe behavioral problems often has worse life outcomes than a stupid person with excellent behavioral strategies that win them the ongoing support and loyalty of those around them, as opposed to antagonizing everyone.

When you're a member of an ethnic group where you are the first person in your family to go to college, it's hard to be sure whether you're first because your ancestors were poor, or whether they were stupid. Since poverty is an obvious explanation, and the null hypothesis is "NO difference between races in basic intelligence," rather than "lots of difference..." I suspect poverty plays a significant role.

Sure, but it's also possible that their poverty was partially caused by their, well, relative stupidity, at least in the generations following the civil rights act and a relatively universal education system. And your g correlates with your future job performance and IIRC your income better than your family's socio-economic status.

Thing is, if SES can entangle with 'g' (say, by depriving children of something they need in early childhood development), then it undermines attempts to establish causation.

It's fun, on some level, to say that people who fail have failed because they are stupid, or to dismiss entire categories of people as stupid and excuse ourselves from any responsibility for their collective failure. But given that this IS an appealing thing psychologically and that it is so easily pointed at, well, pretty much any group one doesn't like... we should inherently be cautious about it.

It's a well document fact that human beings tend to attribute their own failures to extenuating circumstance, and others' failures to innate negative qualities. People actually tend to think:

"If I get upset it's because I had a bad day, if you get upset it's because you're an angry person. If I fire you it's because your performance was inadequate, if you fire me it's because you're a callous bastard. If I flunk out of college it's because the work was too hard and I ran out of money, but if you flunk out of college it's because you're a stupid guy from a stupid family."

As is so often the case when research seems to uphold a deep-rooted psychological bias, I am cautious about embracing the results.

What people justly blame schools and parents for is for failing to make the most of what's available, or for writing children off as 'uneducable' or 'hopeless.' Especially when with discipline and effort and a program oriented toward their talents they could make something useful of themselves. Maybe not in a field requiring great abstract thought and a commanding intellect, but somewhere in some way. If you don't give them what they need, though, they tend to turn into delinquents or useless semi-employable drones.

Sure, but if you want to measure how well you're making do with what's available, you have to control for that, and most commentators aren't willing to.[/quote]

[Ziggy Stardust]

"proving" causation is probably only possible via genetic linking (which has been done to some extent - given time I can pull sources), but it's not that difficult to produce statistics that make causation the most logical answer.

Thus proving you know absolutely nothing about statistics. But hey, if you think really think it's that simple, why don't you tell me what statistics will "make causation the most logical answer"? Also, here's a hint: genetic linking doesn't prove causation, either. If it did we'd have a cure for every genetic disease already.

For one thing, we can look at twin adoption studies, where the adoptee's adult IQ correlates more with that of their biological parents than their adoptive ones - this makes the possibility of socioeconomic and cultural factors being the primary cause of IQ (at least within the spread of 99% of a western democracy) incredibly unlikely. We can then observe the correlation of g with relevant life outcomes while controlling for every other measurable variable, and we find that there remains something substantial. At that point, causation is the most likely explanation.

Nope. This is introductory statistical theory level incorrect. Correlations are still correlations, no matter how many measurable variables you control for. Correlations that have been controlled for other measurable variables are called "partial correlations". And you can't necessarily infer causal pathways using partial correlations, at least not in a complex system with imperfect measurement (which IQ is ... broadly speaking, IQ is a useful short-hand descriptor of certain classes of mental processes. However, it is not by any means widely accepted as an actual universal summary of all types of "intelligence". It is an oversimplification based on largely outdated models that just happens to have some useful properties - and, in fact, it has often been shown that the distribution of IQ is different within different societies/cultures, with biases shown towards Western cultural norms).

[Luke Skywalker]

Even if we assume that your whole argument, that I can't show causation, is correct, it wouldn't get rid of IQ's usefulness as a statistical measure at all.

Thus proving you know absolutely nothing about statistics. But hey, if you think really think it's that simple, why don't you tell me what statistics will "make causation the most logical answer"? Also, here's a hint: genetic linking doesn't prove causation, either. If it did we'd have a cure for every genetic disease already.

Vaguely proclaiming that somebody "doesn't know what they're talking about" and asking them to "figure it out on their own" is the oldest rhetorical trick in the book. It's neither efficient for communicating information nor reasonable to expect me to anticipate your arguments, make them for you, and then refute them on the fly. So what, exactly, is wrong with what I had said?

As for which correlations make causation more likely, see below.

Nope. This is introductory statistical theory level incorrect. Correlations are still correlations, no matter how many measurable variables you control for.

Read what I said more carefully; I didn't claim that you could "prove" correlation statistically. I pointed out that by a certain point, higher IQ causing these other variables to a greater extent than the reverse becomes more likely. Establishing this requires us to inject the context of our variables into the equation beyond pure statistical analysis;

So for example, we know that you can better predict two people's performance on the SAT from their IQ scores than from any other variable we've managed to measure. If IQ scores were just weak proxies for some other variables that were the real cause of this predictive power, you'd expect to get more accurate predictions from just directly measuring said variables. If these hidden factors exist, nobody has yet to find them.

You might then suppose that IQ scores reflect some ideal weighted combination of other factors, and that's why it predicts things so well. But then we can isolate data points where everything else is randomized in whatever manner you want, and IQ still comes out on top. There comes a certain point when the straightforward answer that being smarter helps you in life becomes the most reasonable one.

So we have to basically rule out competing possibilities:

• That IQ is just a reflection of socioeconomic status -> demonstrate superior correlation even when you adjust for other variables.
• That IQ represents some sort of weighted average of said variables -> demonstrate validity independent of all other data, or with twin adoption studies that can arbitrarily vary everything else.
• IQ reflects some measure of work ethic -> attempt to improve IQ in a manner that transfers to an arbitrary battery of tests, and find that this doesn't work.

If you have any other possibilities, please let me know. It's probably already been tested.

BTW, you do realize that I'm not exactly inventing my own notions here, right? This g thing is probably the single most reliable and stable metric that's ever come out of an entire academic field (psychometrics). It's not just my pet theory.

Correlations that have been controlled for other measurable variables are called "partial correlations".

Which make causation more likely...

And you can't necessarily infer causal pathways using partial correlations, at least not in a complex system with imperfect measurement

I don't think you quite understand what IQ is supposed to measure. It isn't universally accepted among the general population, and even among psychologists, but its consensus among psychometricians is fairly unanimous. So:

(which IQ is ... broadly speaking, IQ is a useful short-hand descriptor of certain classes of mental processes.

Incorrect. IQ tests are evaluated on their ability to correlate with the general factor. The general factor is explicitly independent of the specific type of mental process. You're referring to domain specific abilities, which have their own common correlations independent of g, but g correlates with everything.

Your confusion probably comes from noticing that certain tests will, for example, measure your arithmetic, vocab, etc. separately. While this doubles in function for sorting out particular learning disabilities, for the purposes of extracting g, the important part isn't the specific content - it's the general factor that everything correlates to. You measure something like 12 subtests because everyone has domain-specific factors that may make one test slightly (or in some cases highly) higher or lower than others, and if you get a large enough battery of tests you get more statistical significance. Raven's matrices are the best at extracting g and they do this by giving you some patterns and shapes.

It is an oversimplification based on largely outdated models that just happens to have some useful properties - and, in fact, it has often been shown that the distribution of IQ is different within different societies/cultures, with biases shown towards Western cultural norms).

Oh really? I'd love to hear your explanation for how Raven's Progressive Matrices, which are the best at extracting g, can be culturally biased when they're based purely around abstract pictures.

Far from coming from "outdated models", IQ tests have been rigorously refined over time to eliminate all of the allegations that you've thrown at them, which frankly seem to come out of your intuition impression of them rather than the actual data.

[Esquire]

Even if we assume that your whole argument, that I can't show causation, is correct, it wouldn't get rid of IQ's usefulness as a statistical measure at all.

It absolutely would, at least for the purpose you're trying to use IQ for. If IQ is dependent on societal factors - and it is - it's not at all useful as an individual predictor of... well, anything.

So for example, we know that you can better predict two people's performance on the SAT from their IQ scores than from any other variable we've managed to measure. If IQ scores were just weak proxies for some other variables that were the real cause of this predictive power, you'd expect to get more accurate predictions from just directly measuring said variables. If these hidden factors exist, nobody has yet to find them.

You might then suppose that IQ scores reflect some ideal weighted combination of other factors, and that's why it predicts things so well. But then we can isolate data points where everything else is randomized in whatever manner you want, and IQ still comes out on top. There comes a certain point when the straightforward answer that being smarter helps you in life becomes the most reasonable one.

Alternatively, both IQ and SAT scores are both subject to significant confounding from additional variables, invalidating both as a measure of... well, again, anything except themselves. The definition of a confounder is, paraphrased: 'anything that is associated with both exposure and outcome and falls in the causal pathway between the two;' any number of social, economic, or cultural factors fits those criteria.

So we have to basically rule out competing possibilities:

• That IQ is just a reflection of socioeconomic status -> demonstrate superior correlation even when you adjust for other variables.
• That IQ represents some sort of weighted average of said variables -> demonstrate validity independent of all other data, or with twin adoption studies that can arbitrarily vary everything else.
• IQ reflects some measure of work ethic -> attempt to improve IQ in a manner that transfers to an arbitrary battery of tests, and find that this doesn't work.

If you have any other possibilities, please let me know. It's probably already been tested.

BTW, you do realize that I'm not exactly inventing my own notions here, right? This g thing is probably the single most reliable and stable metric that's ever come out of an entire academic field (psychometrics). It's not just my pet theory.

You know, there's a reason why the psychological fields have a reputation as the absolutely least-reliable 'scientific' disciplines. Any field that depends on voting to determine its clinical disorders [see DSM revision procedures] and small samples of psychology undergradutates for its studies isn't doing itself any favors in the credibility department.

Correlations that have been controlled for other measurable variables are called "partial correlations".

Which make causation more likely...

And you can't necessarily infer causal pathways using partial correlations, at least not in a complex system with imperfect measurement

I don't think you quite understand what IQ is supposed to measure. It isn't universally accepted among the general population, and even among psychologists, but its consensus among psychometricians is fairly unanimous. So:

(which IQ is ... broadly speaking, IQ is a useful short-hand descriptor of certain classes of mental processes.

Incorrect. IQ tests are evaluated on their ability to correlate with the general factor. The general factor is explicitly independent of the specific type of mental process. You're referring to domain specific abilities, which have their own common correlations independent of g, but g correlates with everything.

Your confusion probably comes from noticing that certain tests will, for example, measure your arithmetic, vocab, etc. separately. While this doubles in function for sorting out particular learning disabilities, for the purposes of extracting g, the important part isn't the specific content - it's the general factor that everything correlates to. You measure something like 12 subtests because everyone has domain-specific factors that may make one test slightly (or in some cases highly) higher or lower than others, and if you get a large enough battery of tests you get more statistical significance. Raven's matrices are the best at extracting g and they do this by giving you some patterns and shapes.

You realize, of course, that the more tests you run the higher the chance of (literally) purely-coincident positive results? Familywise-alpha is a thing; it's why ANOVA and t-tests are different procedures in the first place.

It is an oversimplification based on largely outdated models that just happens to have some useful properties - and, in fact, it has often been shown that the distribution of IQ is different within different societies/cultures, with biases shown towards Western cultural norms).

Oh really? I'd love to hear your explanation for how Raven's Progressive Matrices, which are the best at extracting g, can be culturally biased when they're based purely around abstract pictures.

Far from coming from "outdated models", IQ tests have been rigorously refined over time to eliminate all of the allegations that you've thrown at them, which frankly seem to come out of your intuition impression of them rather than the actual data.

That's rather a positive claim, don't you think? Please feel free to provide evidence that modern IQ testing has been sufficiently modified to avoid selection, interviewer, and cultural bias; that previously-cited studies are recent enough to have benefited from the above, and that your g-hypothesis satistifies standard causal criteria.

[Luke Skywalker]

Sorry, I've been incredibly busy; I should be able to reply by this week.