Intelligence, Critical Thought Capacity, and Heridability

[69dodge]

I think part of the problem is that I'm combining too many questions about regression to the mean and the heridability of adult IQ into each post.

So let me start with just one:

What exactly is the mean that that the adult IQ of of the offspring of adult IQ deviant parents regress to?

I take it that it's the mean adult IQ of the species as a whole? But why would it be, unless we're keyed to a specific adult IQ as a species, and randomness causes individual adult IQs to fall along a bell curve around it?

No, it's not the mean IQ of the whole species. It's the mean IQ of a (hypothetical, I suppose) bunch of people with the same genes as the parent in question. We don't know for sure what that mean is, but it's probably lower than the parent's IQ, given a random high-IQ parent ("high-IQ" means high relative to the species average, here). drkitten explained it well, I thought.

 

[drkitten]

Let me clarify. Why else would the nonheritable, nonenvironmental (chance elements of how the parent's genes combined in the offspring) regress to the mean IQ of humans, if we weren't genetically keyed to that mean IQ?

Because any statistical distribution has a mean, and any non-pathological distribution will display regression to the mean.

There's no requirement for anything "genetically keyed" to anything, or even to anything "genetic" -- regression to the mean just happens.

Let me frame it differently. Let's say a catastrophe in 2007 wiped out everyone that currently had an adult IQ below 140 as measured in 2006. I understand that in 2007 the new mean would be called 100. But it would still be a population of very deviant adult IQ compared to the world's population in 2006.

All right. Just for the sake of clarity, I will assume consistency of IQ testing (and ignore the well-known Flynn effect) -- so when I talk about IQs in this new, post apocalyptic population, I am referring to IQ scores as tested using 2006 tests and 2006 scaling tables. I will also note that the numbers I'm citing are also pulled out of thin air.

The apocalypse happens, and the mean IQ is now 145.

The "genetic component" of that mean IQ is probably something like 130 -- there are more people who survived the catastrophe who were dumb-but-lucky than there were who were smart-but-unlucky.

Now let's fast forward to the year 2066. The survivors of 2007 have all had kids that have grown up to be adults.

The mean IQ is now 130 in 2006 terms.

For every person who got "lucky" and got an IQ higher than his genetic determination, there is another person who got "unlucky." That's the way that luck distributes. The mean IQ (145) has "regressed" towards the 2006 mean, since 130 < 145.

Having said this, this new 130 mean will be relatively stable; the grandchildren will also have a mean IQ of 130, as will the great-grandchildren, and so forth (discounting the Flynn effect).

If we are keyed to an IQ as a species, and 140 is very deviant from it, then the adult IQ of the entire population would be expected to regress towards the 2006 mean.

The entire population would be expected to regress towards the 2006 mean no matter what. Any distribution will regress toward the mean.

If there is a regression to the 2006 mean in the 2066 offspring population, I think that would indicate that we're genetically keyed to a certain IQ,

No. Any more than the fact that when we fall down, that indicates that we're "genetically keyed" to lower altitudes.

I don't understand logically why there would be a deviation only to the new 2007 mean in the 2066 population, because the 2007 mean would have been very deviant in 2006. Instead, I would think that nonenvironmental luck would make it equally likely for two parents with deviant IQ in 2007 to have an even more deviant offspring as a less deviant offspring.

I still don't understand what you mean by "nonenvironmental luck," and the entire sentence above makes no sense to me whatsoever.

 

[Dave1001]

No, it's not the mean IQ of the whole species. It's the mean IQ of a (hypothetical, I suppose) bunch of people with the same genes as the parent in question. We don't know for sure what that mean is, but it's probably lower than the parent's IQ, given a random high-IQ parent ("high-IQ" means high relative to the species average, here). drkitten explained it well, I thought.

I thought drkitten explained his/her points well too. But I think I've read (both here) and in the wikipedia section on IQ claims that the adult IQ of offspring of deviant adult IQ parents will regress towards the mean of the whole species, rather than specifically to the mean adult IQ of folks with the same genes as the parents. If it regresses to the mean adult IQ of folks with the same genes of the parents then we're just talking about environmental and test-taking randomness. But if it regresses to the mean adult IQ of the species, controlling for environment, then it would still seem to me that would indicate that we're keyed to an adult IQ as a species. Of course, if it's not the mean IQ of the whole species that adult offspring regresses to (perhaps I should say towards), as you say, then that's an easy answer to that question.

In the hypothetical I gave in an earlier post, (where a cataclysm in 2007 wiped out everyone who had below a 140 adult IQ as measured in 2006) according to how you defined regression to the mean and how it plays out in heredability and adult IQ, what would expect the adult offspring's IQ to regress to in 2066? About a 130-135 in 2006 measures? Because if as others have posted in here there's about a 10 point variation due to testtaking factors, then probably that's where they'd fall. Maybe somewhat lower factoring other environmental randomness such as womb environment. So maybe they'd regress as a population to a 120 in 2006 measures? And then that would remain the mean IQ of the 2066 population, and their offspring, and their grandoffspring, measured against the 2006 curve?

 

[drkitten]

No, it's not the mean IQ of the whole species. It's the mean IQ of a (hypothetical, I suppose) bunch of people with the same genes as the parent in question.

Well, in strict terms, it's the mean of whatever distribution we originally sampled from.

"Mean" only makes sense relative to a distribution; if I'm only looking at a single data point, there's no meaning we can discuss for "regression to the mean."

In order to intelligently discuss "regression to the mean," we need at least three things. FIrst, a large population with well-defined characteristics. Second, a well-defined sample that differs in some way from the population as a whole. (I.e., my sample is "all of humanity," my well-defined sample is "all of humanity with an IQ of 130 or better as measured on an IQ test given 24/6/2006"). Third, we need a second, related sample -- ideally with paired data. This could be "the children of the sample," or it could simply be "the test scores of the sample as obtained on 1/8/2006."

Regression to the mean simply notes that the mean of the third group (as measured on the property of interest) will be between the universal mean and the mean of the second group..

 

[Dave1001]

Well, in strict terms, it's the mean of whatever distribution we originally sampled from.

"Mean" only makes sense relative to a distribution; if I'm only looking at a single data point, there's no meaning we can discuss for "regression to the mean."

In order to intelligently discuss "regression to the mean," we need at least three things. FIrst, a large population with well-defined characteristics. Second, a well-defined sample that differs in some way from the population as a whole. (I.e., my sample is "all of humanity," my well-defined sample is "all of humanity with an IQ of 130 or better as measured on an IQ test given 24/6/2006"). Third, we need a second, related sample -- ideally with paired data. This could be "the children of the sample," or it could simply be "the test scores of the sample as obtained on 1/8/2006."

Regression to the mean simply notes that the mean of the third group (as measured on the property of interest) will be between the universal mean and the mean of the second group..

drkitten. Okay, let's go with my original example, but instead of everyone with below 140 IQ being wiped out, folks with 140 and higher IQ as measured 24/6/2006 are simply sequestered in a biodome on 1/1/2007.

I understand that the sequestered population's offspring will have an adult IQ between 100 (the universal mean) and 140 (the mean of the second group) by the standards of the 24/6/2006 test as measured on 1/1/2066. Now according to what your saying here, if I also retested the parents on 1/1/2066 they'd also have the same, lower IQ as their kids than how they scored on 24/6/2006? Cause from other posts here and from wikipedia, I think the much much older now adults would score below their 140 (they'd probably score a 130-135) but their offspring would score even lower than them (maybe 120-125?), because their offspring would also probably be less lucky than them in womb environment and developmental environment (and yes, maybe even the way their genes combined -but please don't get sidetracked by this literally parenthetical sidepoint that everyone seems to disagree with).

However, their offspring, tested in 1/1/2066 have now had what is presumed to be an average amount of luck and have ended up with adult IQ's about 120-125. Should we expect the adult grandoffspring, still living in the biodome, tested in 1/1/2140 to also have a 120-125 adult IQ as measured against the 24/6/2006 entire species population that took the test? Why would they revert to the "universal mean" further than did the new adult population of 1/1/2066? I take it that both you and 69dodge think they wouldn't continue to revert further toward the 100 mean IQ of the the entire population tested on 24/6/2006.

 

[drkitten]

I thought drkitten explained his points well too. But I think I've read (both here) and in the wikipedia section on IQ claims that the adult IQ of offspring of deviant adult IQ parents will regress towards the mean of the whole species, rather than specifically to the mean adult IQ of folks with the same genes as the parents. If it regresses to the mean adult IQ of folks with the same genes of the parents then we're just talking about environmental and test-taking randomness. But if it regresses to the mean adult IQ of the species, controlling for environment, then it would still seem to me that would indicate that we're keyed to an adult IQ as a species.

I don't understand what the highlighted phrase means in this context. If you're controlling for environment in one statistical study, but not in another, then you're comparing two different things and you should not expect the same answer.

In the hypothetical I gave in an earlier post, (where a cataclysm in 2007 wiped out everyone who had below a 140 adult IQ as measured in 2006) according to how you defined regression to the mean and how it plays out in heredability and adult IQ, what would expect the adult offspring's IQ to regress to in 2066? About a 130-135 in 2006 measures? Because if as others have posted in here there's about a 10 point variation due to testtaking factors, then probably that's where they'd fall.

That's part of the problem -- you're not being clear about what factors you are and are not considering as "environmental" and "genetic."

There are at least four distinct attributes involved.

1) Genetic stuff. This is the stuff that I would share with my biological clone, probably not directly testable.
2) Prenatal environment. This, plus the above, is what I would share with my separated-at-birth identical twin. Separated twin studies could therefore give us a handle on the significance of these factors -- what's the probability of my IQ being x, given that my separated twin has an IQ of y{/i]?
3) Developmental environment. This is the long-term stuff that is consistent to me, but I don't share with my separated twin. I had a really bad second-grade teacher, but a really good track coach. I grew up in an English-speaking environment, surrounded by lots of books. Et cetera. Again, twin studies might help here.
4) Test-taking factors. This is the short-term stuff that is not consistent to me. I got a good breakfast the last time I took an IQ test, but not this one. This is the reason I can take the same IQ test four times and get four different scores.

Which of those are you considering "environmental"? Because you'll see "regression to the mean" at any of those levels.

Suppose some mad scientist gives everyone in the world an IQ test on August 1, 2006 -- and kills everyone who doesn't score 130 exactly. The world mean score is of course 100 -- the mean of the subset remaining is of course 130 by definition.

What happens on August 2, 2006, when he re-tests everyone? Will everyone get the same score? Of course not. Test-taking factors will be different. Most of the group will probably score between 120 and 140 on that test, but we'll see outliers all over the map.

What will be the survivor's mean score? Not 130. Lower than 130, and closer to the "universal" mean of 100. That's "regression to the mean." With nothing "genetic" involved.
If we re-test often enough, we will find that our group of survivors probably has a "real" mean IQ closer to, say 125.

 

[Bruno Putzeys]

I would like to point out that the IQ in a population that's chopped off below 140 is no longer normally distributed.

 

[drkitten]

drkitten. Okay, let's go with my original example, but instead of everyone with below 140 IQ being wiped out, folks with 140 and higher IQ as measured 24/6/2006 are simply sequestered in a biodome on 1/1/2007.

BIg biodome.

I understand that the sequestered population's offspring will have an adult IQ between 100 (the universal mean) and 140 (the mean of the second group) by the standards of the 24/6/2006 test as measured on 1/1/2066. Now according to what your saying here, if I also retested the parents on 1/1/2066 they'd also have the same, lower IQ as their kids than how they scored on 24/6/2006?

No, they wouldn't have the same, lower, IQ. But they'd have a lower mean IQ than they got on 24/6/2006. Because the biodome internees will tend to be "lucky" in their test taking factors, a luck that doesn't carry over to the next test.

Cause from other posts here and from wikipedia, I think the much much older now adults would score below their 140 (they'd probably score a 130-135) but their offspring would score even lower than them (maybe 120-125?), because their offspring would also probably be less lucky than them in womb environment and developmental environment .

Yes. The biodome internees will tend to be "lucky" in their developmental environment, a luck that doesn't carry over to the next generation. And furthemore, their children will not be especially "lucky" in their testing environment, which again lowers their score. So the children's mean score will be lower than the parents' mean score on the retest, which in turn will be lower than the parents' mean score on the selection test.

However, their offspring, tested in 1/1/2066 have now had what is presumed to be an average amount of luck and have ended up with adult IQ's about 120-125. Should we expect the adult grandoffspring, still living in the biodome, tested in 1/1/2140 to also have a 120-125 adult IQ as measured against the 24/6/2006 entire species population that took the test?

There is no reason to assume that the grand-offspring would revert further to the mean, all else being equal.

Basically, the offspring are a representative sample (*) of the gene pool of the internees.
The grand-offspring are also a representative sample of the gene pool of the internees.
Two representative samples are expected to have the same mean, plus or minus random error factors.

(*) It's the "representative sample" that's the key issue. The internees are not a representative sample of the human population at large, since they were selected unrepresentatively. From a biogenetic standpoint, the idea that children are a "representative sample" of their parents' gene pool is somewhat questionable, both due to things like natural selection and other aspects like assortative mating. If you start looking in detail at this level of biology, you're no longer dealing with a purely statistical issue and the math starts to get really difficult.

 

[Dave1001]

I don't understand what the highlighted phrase means in this context. If you're controlling for environment in one statistical study, but not in another, then you're comparing two different things and you should not expect the same answer.

I meant controlling for the environment in every statistical study, because environment seems to me to generally be keyed to the population mean, particularly for offspring of negatively adult IQ deviant parents. We have social mechanisms like prenatal health laws and resources, public schools, and pop culture that probably ensures that offspring of negatively adult IQ deviant parents grow up in pretty similar cognitive development environments as the offspring of adult IQ average parents. With adult IQ positively deviant parents, it's probably trickier, because to a degree they may substantially enrich the developmental environment for their offspring, from prenatal through devlopmental years, but still, their offspring probably spend a substantial amount of time in the same developmental environments as the offspring of adult IQ average parents too.

The reason I brought up controlling for environment is because I was curious the degree to which nonheritable genetic luck would result in reversion to the mean. It looks like based on my hypothetical (the one with the sequestered high adult IQ subset of the population) that folks don't think that genetic luck plays a role, at least genetic luck at being born with nonheritable genes very different from the genetic makeup of an adult IQ average person (for the hypothetical that would be an adult IQ average person that took the IQ test in 2006).

That's part of the problem -- you're not being clear about what factors you are and are not considering as "environmental" and "genetic."

There are at least four distinct attributes involved.

1) Genetic stuff. This is the stuff that I would share with my biological clone, probably not directly testable.
2) Prenatal environment. This, plus the above, is what I would share with my separated-at-birth identical twin. Separated twin studies could therefore give us a handle on the significance of these factors -- what's the probability of my IQ being x, given that my separated twin has an IQ of y{/i]?
3) Developmental environment. This is the long-term stuff that is consistent to me, but I don't share with my separated twin. I had a really bad second-grade teacher, but a really good track coach. I grew up in an English-speaking environment, surrounded by lots of books. Et cetera. Again, twin studies might help here.
4) Test-taking factors. This is the short-term stuff that is not consistent to me. I got a good breakfast the last time I took an IQ test, but not this one. This is the reason I can take the same IQ test four times and get four different scores.

Which of those are you considering "environmental"? Because you'll see "regression to the mean" at any of those levels.

As I think I've said a couple times before now, I consider all of those factors environmental. Basically, any factor that's not genetic (either heritably genetic or randomly genetic) I'm calling environmental here. I hope this clears this up completely for you.

Suppose some mad scientist gives everyone in the world an IQ test on August 1, 2006 -- and kills everyone who doesn't score 130 exactly. The world mean score is of course 100 -- the mean of the subset remaining is of course 130 by definition.

What happens on August 2, 2006, when he re-tests everyone? Will everyone get the same score? Of course not. Test-taking factors will be different. Most of the group will probably score between 120 and 140 on that test, but we'll see outliers all over the map.

What will be the survivor's mean score? Not 130. Lower than 130, and closer to the "universal" mean of 100. That's "regression to the mean." With nothing "genetic" involved.
If we re-test often enough, we will find that our group of survivors probably has a "real" mean IQ closer to, say 125.

Got that. I'm pretty sure you posted this point already. I'm looking forward to your answers and to my follow up questions on this.

 

[Dave1001]

BIg biodome.




No, they wouldn't have the same, lower, IQ. But they'd have a lower mean IQ than they got on 24/6/2006. Because the biodome internees will tend to be "lucky" in their test taking factors, a luck that doesn't carry over to the next test.



Yes. The biodome internees will tend to be "lucky" in their developmental environment, a luck that doesn't carry over to the next generation. And furthemore, their children will not be especially "lucky" in their testing environment, which again lowers their score. So the children's mean score will be lower than the parents' mean score on the retest, which in turn will be lower than the parents' mean score on the selection test.




There is no reason to assume that the grand-offspring would revert further to the mean, all else being equal.

Basically, the offspring are a representative sample (*) of the gene pool of the internees.
The grand-offspring are also a representative sample of the gene pool of the internees.
Two representative samples are expected to have the same mean, plus or minus random error factors.

(*) It's the "representative sample" that's the key issue. The internees are not a representative sample of the human population at large, since they were selected unrepresentatively. From a biogenetic standpoint, the idea that children are a "representative sample" of their parents' gene pool is somewhat questionable, both due to things like natural selection and other aspects like assortative mating. If you start looking in detail at this level of biology, you're no longer dealing with a purely statistical issue and the math starts to get really difficult.

drkitten, fantastic answers to my questions. Thank you.

Now, is there anyone who disagrees with drkitten or thinks drkitten is in error on any aspect of this analysis? If so, please give a detailed critique and explain why you think the results from my hypothetical would be different than what he/she has laid out.

 

[hammegk]

**I recently was debating a neuroscientist on this, who pointed my to a study suggesting that h2 estimates are flawed because they ignore the prenatal environment.

Her argument was it's the womb that matters. Clearly, this is an environmental variable, ....

Environmental perhaps, but would not one expect a strong correlation between mom's IQ (dad's too perhaps) and quality of the prenatal environment she provides? Why would anyone argue that genetic heritage of ancestors stops at conception?

 

[Ben Tilly]

Environmental perhaps, but would not one expect a strong correlation between mom's IQ (dad's too perhaps) and quality of the prenatal environment she provides? Why would anyone argue that genetic heritage of ancestors stops at conception?

It gets more complex than that. For instance the effects of genetic heritage can operate through environmental factors.

For instance consider otherwise equivalent babies, but one is ugly and the other is attractive. The attractive kid will, all else being equal, get more attention, and more positive attention. This results in an environmental boost that is likely to make that kid more intelligent. (Note that there is a known positive correlation between attractiveness and intelligence.) So genes causing babies to be attractive result in them being intelligent.

Incidentally in twin studies of the heritability of intelligence, effects like this will show up as intelligence being inheritable. Why? Because effects like this cause identical twins to have a more similar environment than fraternal twins, and therefore IQs of identical twins are more similar than of fraternal twins. And yes, there is an inherited characteristic affecting intelligence - but when you understand the mechanism it doesn't support the idea that your intelligence is preordained by your genetics.

Cheers,
Ben

 

[Dave1001]

It gets more complex than that. For instance the effects of genetic heritage can operate through environmental factors.

For instance consider otherwise equivalent babies, but one is ugly and the other is attractive. The attractive kid will, all else being equal, get more attention, and more positive attention. This results in an environmental boost that is likely to make that kid more intelligent. (Note that there is a known positive correlation between attractiveness and intelligence.) So genes causing babies to be attractive result in them being intelligent.

Incidentally in twin studies of the heritability of intelligence, effects like this will show up as intelligence being inheritable. Why? Because effects like this cause identical twins to have a more similar environment than fraternal twins, and therefore IQs of identical twins are more similar than of fraternal twins. And yes, there is an inherited characteristic affecting intelligence - but when you understand the mechanism it doesn't support the idea that your intelligence is preordained by your genetics.

Cheers,
Ben

There are alternative explanations to consider, like that both intelligence and attractiveness are inheritable, and relatively intelligent people have been mating with the relatively attractive people (facial symmetry, etc.) for many generations.

 

[hammegk]

I'm tending towards the idea that at conception, all maximum potential is coded in. All environment can do is not allow it to be reached, and it's a lucky individual indeed who receives nothing but 'good environment'.

 

[Ben Tilly]

There are alternative explanations to consider, like that both intelligence and attractiveness are inheritable, and relatively intelligent people have been mating with the relatively attractive people (facial symmetry, etc.) for many generations.

Alternate explanations of what?

Did you, perchance, think that I was saying that the correlation between attractiveness and intelligence is entirely explained by one environmental effect?

If so, then you're very mistaken. I believe that there is no single explanation of it. Instead there are many contributing factors to this correlation. I mentioned one thing that I believe contributes. There are others. For a random instance, some of the same environmental factors that contribute to intelligence (eg good nutrition, not being too sick, etc), also contribute to attractiveness.

Cheers,
Ben

 

[Dave1001]

I believe that there is no single explanation of it. Instead there are many contributing factors to this correlation. I mentioned one thing that I believe contributes. There are others. For a random instance, some of the same environmental factors that contribute to intelligence (eg good nutrition, not being too sick, etc), also contribute to attractiveness.

Cheers,
Ben

that seems reasonable.