Turkheimer and Waldron: When genetic similarity is controlled, siblings often appear no more alike than individuals selected at random from the population. ... it has become widely accepted that the source of this dissimilarity is a variance component called nonshared environment.
... In what may have been the most influential article ever written in the field of developmental behavior genetics, Plomin and Daniels (1987) reviewed evidence that a substantial portion of the variability in behavioral outcomes could not be explained by the additive effects of genotype or the environmental influences of families. They suggested that this residual term, which they called the nonshared environment, had been neglected by environmentally oriented researchers who assumed that the most important mechanisms of environmental action involved familial variables, like socioeconomic status [SES] and parenting styles, that are shared by siblings raised in the same home and serve to make siblings more similar to each other. Indeed, Plomin and Daniels argued, once genetic relatedness has been taken into account, siblings seem to be hardly more similar than children chosen at random from the population.
In other words, despite a lifetime of proximity, your adopted child may bear no more similarity to you (in terms of, e.g., intelligence) than someone selected at random from the general population. The shared family environment that your children (biological or adopted) experience has little or no measurable effect on their cognitive development. While there are environmental effects on intelligence (the highest estimates of heritability for adult IQ are around .8, and some would argue for a lower value; see here for Turkheimer's work suggesting low heritability in the case of severe deprivation), they seem to be idiosyncratic factors that can't be characterized using observable parameters such as the parents' SES, parenting style, level of education, or IQ. It is as if each child experiences their own random micro-environment, independent of these parental or family characteristics.
The nonshared influences are by far the largest environmental (non-genetic) influences on intelligence -- in fact, they are the only detectable non-genetic influences. (Click figure for larger version; from a review by Plomin. More recent overview here.)
Identical twins, whether raised together or apart, turn out to be very similar, but one still finds differences in IQ and personality. The cause of those differences must be the different environments experienced by the twins, but can't be characterized by simple variables of the sort listed above: it is not the case that the twin raised by the higher SES family has, on average, a much higher IQ! In fact, twins raised in the same family are about as similar as those raised apart, so family shared environment does not produce a large measurable influence. See below for a plausible model that accounts for such outcomes.
By now these results are well understood and accepted by experts, but not by the general population or even policy makers. (See the work of Judith Rich Harris for popular exposition). The naive and still widely held expectation is that, e.g., high SES causes a good learning environment, leading to positive outcomes for children raised in such environments. However, the data suggests that what is really being passed on to the children is the genes of the parent, which are mainly responsible for, e.g., above average IQ outcomes in high SES homes (surprise! high SES parents actually have better genes, on average). Little or no positive effect can be traced to the SES variable for adopted children.
The implications are quite shocking, especially for two groups: high investment parents (because the ability of parents to influence their child's development appears limited) and egalitarians (because the importance of genes and the difficulty in controlling environmental effects seem to support the Social Darwinist position widely held in the previous century).
It is plausible to me that each child tends to create their own environment over time, by selectively seeking out or avoiding stimuli of various types. A bookish kid may end up at the library regardless of whether their father takes them there. An athletic kid may end up on the playground whether or not their mother takes them there. It has been argued that this effect is the reason that the heritability of IQ increases with age: over time, genetic influences assume greater importance as they cause the individual to create or seek out their preferred environment.
In a previous post I discussed individual cognitive profiles as described by an n-vector. Similarly, one could think of an individual's learning profile and learning environment as two more n-vectors. These n-vectors may or may not be well-matched, leading to outcomes with significant and hard to characterize variability. For example, one can imagine that both the environment (provided by parents, siblings, teachers and peers) and a particular child's reactions vary in each of the factors listed below.
Pressure and competition
Stimulation through stories and pretend play; flights of imagination
Ability to learn from repetition and drill / tendency to boredom
Isolated study vs group activities
Visual vs aural vs mechanical stimulation
Level of discipline or structure imposed
Close mentoring vs freedom of exploration
Abstraction vs experimentation
(One can think of many more.)
The factors listed are not intrinsically good or bad for learning -- what matters is whether the learning environment is matched to the nature of the individual child. Some react well to discipline or pressure or story telling, others do not. Further, none of the factors is obviously correlated with SES, parental education level or IQ. Even if they were, it's plausible that a child to some extent creates their learning environment outside the control of parents and teachers (e.g., through peer group or choice of play activities).
An individual whose learning vector (learning style) is well matched to their environment will thrive: the nonshared environmental component in their development will be large and positive. For others, the environment will have a smaller or even negative impact. Because both the learning vector and the environment vector vary in a many-dimensional space, and over time, prediction or control of the overall environmental effect on development is difficult.
Nonshared environmental contributions to development, which are the largest environmental contributions, are effectively random. They are not amenable to control, either by parents or policy makers. Note, this picture -- that each child creates their own environment, or experiences an effectively random one -- does not seem to support the hypothesis that observed group differences in cognitive ability are primarily of non-genetic origin. Nor does it suggest that any simple intervention (for example, equalizing average SES levels) will eliminate group differences. However, it's fair to say our understanding of these complex questions is limited.
Technical remark: if n is large, and factors uncorrelated, the observed environmental variation in a population will be suppressed as n^{-1/2} relative to the maximum environmental effect. That means that the best or worst case scenarios for environmental effect, although hard to achieve, could be surprisingly large. In other words, if the environment is perfectly suited to the child, there could be an anomalously large non-genetic effect, relative to the variance observed in the population as a whole. Of course, for large n these perfect conditions are also harder to arrange. (As a super-high investment parent I am actually involved in attempting to fine tune n-vectors ;-)
Environmental effects cause regression to the mean of a child relative to the parental midpoint. Parents who are well above average likely benefited from a good match between their environment and individual proclivities, as well as from good genes. This match is difficult to replicate for their children -- only genes are passed on with certainty.
123 comments:
There is no single heritability. The extent of heritability depends on the range of environments. This is like the corelation of SAT with grades at an elite school with a restricted range of SAT scores.
Once again Steve you have proved that you do not know what you are talking about.
The difference needn't be extreme, whatever "extreme" means. Those like Murray, who despite caveats, continue to speak of heritability as a single number are ridiculous.
***There is no single heritability. The extent of heritability depends on the range of environments.***
That is what Steve said here isn't it?
"the highest estimates of heritability for adult IQ are around .8, and some would argue for a lower value; see here for Turkheimer's work suggesting extremely low heritability in the case of severe environmental deprivation"
Recent paper by Burdick mentions this:
***Data from more than 8000 parent-offspring pairs, 25,000 sibling pairs, 10,000 twin pairs,
and adoption studies provide evidence that genetic factors play a substantial role in the variation
of g (6), with heritability estimates ranging from 40% to 80%. This suggests that there are
genetically-influenced mechanisms that affect performance across a number of diverse cognitive
measures.***
http://hmg.oxfordjournals.org/cgi/reprint/ddi481v1.pdf
No. It isn't what he said. Steve, Murray, and others play a game.
The game is there is one heritability, and that the apparent exceptions are only for "severe deprivation".
Where is the line between severe and not severe? Heritability varies from 0 to as much as .9 with SES, geographic proximity of twins, etc.
Is there some leap from 0 to .9?
Of course not. It must be continuous.
Steve, Jensen, Murray etc. are guilty of reification.
g has about as much reality as the id or superego.
Steve has shown himself susceptible to this eror in his posts on evolution and autism.
Psychology and psychiatry are pseudosciences carried on by those lacking the ability to do real science.
Because there is a word does not mean there is a thing.
Because a sentence is grammatical does not mean it says something.
***Steve, Jensen, Murray etc. are guilty of reification.
g has about as much reality as the id or superego.***
Here is Jensen's response to Gould's accusation of reification.
""reification" is neither more nor less than the common practice in every science of hypothesizing explanatory models or theories to account for the observed relationships within a given domain. Well- known examples include the heliocentric theory of planetary motion, the Bohr atom, the electromagnetic field, the kinetic theory of gases, gravitation, quarks, Mendelian genes, mass, velocity, and so forth. None of these constructs exists as a palpable entity occupying physical space.
The g factor, and theories attempting to explain g in terms of models independent of factor analysis itself, are essentially no different from the other constructs of science listed above. Nor is there any good reason that hypothetical models attempting to account for g should necessarily exclude all considerations of neural or biochemical processes. All such theoretical speculations about the nature of g, whether offered by Spearman, Burt, Jensen, or anyone else, have involved hypothetical processes or system concepts, presumably going on in the brain (where else?). But these theories have never depicted g as some "single," "ineluctable," "hard," "object," of the sort characterized by Gould. Would Gould then deny psychology the common right of every science to the use of hypothetical constructs or any theoretical speculation concerning causal explanations of its observable phenomena? He writes," My complaint lies with the practice of assuming that the mere existence of a factor, in itself, provides a license for causal speculation" (p.268). But haven't all sciences always exercised free license for theoretical speculation about the causes of the observable phenomena in their domains? Of course they have.
The crucial question, which is summarized by the existence of the g factor is this: In respect to what processes or mechanisms is it that persons who perform well on anyone test, in general, also perform well on many other tests, even on tests that are highly dissimilar in content and sensory and motor modalities? The concept of intelligence depends not on the fact that people can be ranked by this test or that, but rather on the fact that, whatever the test, so long as it is cognitive in the broadest sense, a positive correlation emerges between the ranks for any two tests. If an IQ test were just a rag- bag collection of cognitive tasks that did not all measure a common factor, there could be no positive manifold. Scientists today are trying to understand the causes of positive manifold, and this is what the present g theory is all about. Gould offers no alternative ideas to account for all these well- established observations. His mission in this area appears entirely nihilistic."
http://www.debunker.com/texts/jensen.html
***g has about as much reality as the id or superego.***
Brain imaging is making some progress in identifying what g is.
www.loni.ucla.edu/~thompson/PDF/nrn0604-GrayThompson.pdf
http://www.jneurosci.org/cgi/content/abstract/28/41/10323
There are some SES differences in IQ heritability measurements but I would not fall for theTurkheimer deprivation model.
First I expect there is a lot more intense assortative mating for IQ in some groups than in others.
Second I expect that paternity error varies a lot by SES: this will cut IQ sib correlations a lot.
Is the "non-shared environment" tag any more than a cute word trick? Isn't it just what we used to call "error?"
Just on the Turkheimer study, some comments Bob W has made previously.
"SOCIOECONOMIC STATUS MODIFIES HERITABILITY OF IQ IN YOUNG CHILDREN
Eric Turkheimer, Andreana Haley, Mary Waldron, Brian D'Onofrio, and Irving I. Gottesman
"Scores on the Wechsler Intelligence Scale for Children were analyzed in a sample of 7-year-old twins from the National Col-laborative Perinatal Project.
The models suggest that in impoverished families, 60% of the variance in IQ is accounted for by the shared environment, and the contribution of genes is close to zero; in affluent families, the result is almost exactly the reverse."
Here are my comments:
1 - The study included only young children and does not make any attempt to extrapolate that all other findings of significant increases in h^2 by age 17 are in any way invalid. The effects of the shared environment vanish at around age 12.
2 - Turkheimer began his paper by recognizing that the heritability of cognitive ability in childhood is well established.
3 - Turkheimer made no attempt whatsoever to determine what components of SES he was measuring. There are three obvious items to consider: macro environmental, micro environmental, and genetic. All work to date indicates that the first of these can be found in children, but that it is absent in late adolescents; by late adolescence, all of the environmental component is of the second type; and that genetic intelligence is the largest determinant of SES.
4 - Turkheimer says that the effect he observed was related to the homes in which the children were raised. This is interesting, since it relates to the adoption studies which show that after childhood there is no adult IQ correlation between biologically unrelated children who were reared together in the same home.
5 - Turkheimer discusses in some detail that SES is not strictly an environmental variable, since it is known to be (statistically) caused by the intelligence of the parents. He points out that the models he used "cannot determine which aspect of SES is responsible for the interactions" observed.
As for the reason the shared environmental component vanishes there are a number of possible reasons that are not yet sufficiently investigated to establish proof of causation. One is Thompson's finding that the cerebral cortex thickens and then thins over childhood and follows thickness trajectories that are different for different IQ levels. The process is more extended in bright children. This trajectory is established at a very early age. A recent replication of the study was done by Karama, et. al.
When children are the subject, mental age is a factor. It has been known to all generations that children develop their mental capacity as a function of time. The idea is that the genetic expression has not yet taken what amounts to complete control of g in the individual until about age 16, which is usually taken as the equilibrium adult mental age and is used to estimate the relative level of function of low IQ adults.
Some genes do not express themselves immediately. Height, for example, follows the same path as IQ. The heritability of both is low in youth and increases to very high within a few years. This presumably relates to the variation in biological development. In the case of intelligence, h^2 is around 45% in childhood; 70% in early adulthood; and 80-85% in later adulthood. Recent papers showing h^2 calculations have consistently found values in the range of 80 to 85%.
Lastly, the importance of the buildup of myelin has finally been demonstrated via imaging technology. 15 years ago Ed Miller wrote a very thoughtful paper pointing to the increase in myelin as the underlying cause of the variance in intelligence. There is some evidence to the effect that this may vary by sex (my discussions with Haier). In any case myelination is probably responsible for a significant part of the erasure of the shared environmental component, as it builds up at the rate that corresponds to the vanishing of the shared environmental component."
I don't have any specific comments on Turkheimer's heritability results. This post is mainly about the fact that we cannot characterize the non-genetic part of variation using coarse variables like SES, parenting style, etc.
Whether the heritability of IQ is .8 or .5, there is some residual component that isn't explained by genes. What is interesting is that this residual component is (as far as the experiments can detect) completely random -- it doesn't correlate with any variables the experimenters have control over. Before the results of Plomin et al., people thought SES would explain most of this extra variance; if not SES, then parenting style or parental education... but nothing works.
A radical proposal (e.g., Judith Rich Harris) is that each child creates their own micro environment (e.g., by choice of peer group), regardless of family SES or other factors, and this causes the extra random variance. Another possibility is the one I discussed in some detail -- there are parameters characterizing the learner and the learning environment that are more or less complementary. The goodness of this match is not a function of SES or other coarse observables.
Re: anon,
1. I don't reify g. It should be clear from the compression post that I think it's a crude construct, but nevertheless might have some predictive value. My main interest is understanding what the data says.
2. I appreciate that heritability is only defined relative to a specific range of environments. Adoption data is probably limited to working class and higher SES; we might see lower h if we expanded the range of environments (see Turkheimer). But we never see heritability of 1, and pretty much everything that remains is nonshared evironmental influence -- there is no "shared" environment that experiments can detect. This post is about trying to understand that source of variance.
Why don't identical twins create the same environment for themselves?
Do good parents make a differnce? There are bad rich parents, and good poor parents.
My experience is that the personal investment of the parents in their children's lives makes an enormous difference in social outcomes.
"Why don't identical twins create the same environment for themselves?"
Because they don't have complete control over it. They often end up with different math teachers, different best friends, different girlfriends, different life traumas, etc.
"My experience is that the personal investment of the parents in their children's lives makes an enormous difference in social outcomes."
I'm not sure what you mean by social outcomes, but there is no evidence that parental SES or parenting style (more involved / less involved) raises the IQ of the kids. By no evidence I mean the effect is not statistically significant in the crude studies that have been done, which at least limits the size of the effect. I do think a parent can help a kid get their first job, first mortgage; can force the kid to learn certain skills like swimming or boxing, etc. I would guess transmission of cultural values sort of works, but the personality results are like the IQ results -- mostly nonshared variance.
Hey all,
I have a post in the works on my blog about JR Harris's book. Here are some problems with this particular reading of the behavior genetic data:
1. Shared environment matters for a long time. It is not until late adulthood that it fades away.
2. As anon said, there is no single heritability. Lower SES people have lower heritabilities. Ben017 pointed out that Turkheimer (2003) relied on children, which is true, but he ignores that a replication by Turkheimer found that heritability was smaller in lower SES 17 year olds.
3. This whole discussion is neglecting the role of gene-environment interactions, which load onto the heritability quotient.
Sorry if you are planning to cover these points in your post (put the link up when you are done), but...
"1. Shared environment matters for a long time. It is not until late adulthood that it fades away."
Details please?
"2. As anon said, there is no single heritability."
For the millionth time I acknowledged this and my post doesn't rely on any particular value of h.
"3. This whole discussion is neglecting the role of gene-environment interactions, which load onto the heritability quotient."
The learning vector vs environment vector discussion *is* about gene-environment interaction. What causes a child's learning profile? Read it again and think about it.
Henry, I have never heard these criticisms of Turkheimer's work before, very interesting.
First I expect there is a lot more intense assortative mating for IQ in some groups than in others.
Is there evidence for this? Also, even if this is true, it only counters Turkheimer's interpretation, not his finding that genes matter less for the poor.
Second I expect that paternity error varies a lot by SES: this will cut IQ sib correlations a lot.
If a twin study is being done, as Turkheimer did, sib correlations won't be effected by paternity. So I don't see how this could effect results.
Steve,
I see you do acknowledge the idea of gene by environment interaction in the sense of people being matched to the right learning environment. Why do you rule out a gene environment interaction between SES or other shared family factors, though?
As for my claim about the shared environment mattering up until late adulthood, I'm going off of two things: First, Turkheimer (2007) which I mentioned earlier. Second, my copy of Flynn's "What is Intelligence" cites Jensen's "The g Factor" in saying that shared environment evaporates by adulthood. In contrast, I commonly see people referencing the adoption literature to say that the shared environment stops mattering by age 12. I'll admit that I'm not sure what to go with here, if anyone could clarify the diverging results it would be helpful..
ben g,
Just one quick point.
I looked at the paper you referenced regarding Turkenheimer's more recent results on 17 year olds.
Thing is, those results are predicated on data from the old NMSQT.
As Turkenheimer's paper itself says,
"As mentioned previously, NMSQT aptitude scores are not synonymous with IQ scores. The shared environment may be more influential for NMSQT scores because they also reflect, to some extent, academic achievement, rather than pure cognitive ability."
Certainly we expect achievement is affected much more by the shared environment of family values and intensity of school preparation than is pure cognitive ability. The degree such environments are shared among those in a fixed SES rank may well vary considerably across those ranks. Likely though toward the top end of SES the effect of better family values and school quality tends to decrease in importance, virtually all such students having enough to reach their potential. This would emphasize the heritability component.
the idea the achievement and "pure cogntive ability" can be separated is a pipe dream.
the nmsqt is an iq test whether iq experts want it to be or not.
anon,
I have no doubt but that any achievement test, including the old NMSQT and the College Board Achievement tests, are pretty well correlated with IQ tests, and have a pretty strong "g" component.
But that g component mostly is measured when restricting the range to those already of sufficient academic achievement for the test in question. This, for example, is the likely significance of the MCAT, which requires considerable preparation in various subjects even to be able to understand the questions, but which may well mostly test g beyond a baseline of preparation.
In short, what's pretty clear is that for most achievement tests, the overall quality of school preparation -- or lack thereof -- matters, especially in the upper ranges of achievement targeted by the old NMSQT. Schools without advanced curriculum may prepare even very smart students quite inadequately for such a test. Insofar as one is talking about an achievement test such as the old NMSQT which the great majority of college bound student might take, level of general preparation may make a considerable difference. And unlike, say, College Board achievement tests or AP tests, those taking it are not a self selected group of those who have already taken a course in the discipline in question.
The underlying numbers here matter. If overall academic achievement in HS is correlated with the type of school one goes to and the family values one has picked up even after subtracting out the effect of pure cognitive ability, then one might well expect the amount of heritability on that test to vary depending on one's SES.
Certainly the numbers might be affected enough that the impact of shared environment might itself vary with SES.
Steve,
Wouldn't SES be correlated with ability to control one's environment? In partcular, low SES may make it difficult for genetically bookish kid to go to the library. Perhaps, the nonshared environment is due to truly random events at the cellular and macroscopic level.
You don't understand.
What makes an IQ test? Do all self-described IQ tests give the same result? Do they correlate with each other any better than they do with the nmsqt and other such tests?
If a test correlates as well with self-described IQ tests as those tests corrlate with each other why is it not an IQ test?
Marketing?
Anon, such a test has probably been sought. Had it been found I think we would know about it. We cannot prove for absolute certain that it has ever been reported, but no one seems to have heard of it.
"Why do you rule out a gene environment interaction between SES or other shared family factors, though?"
This would show up as a correlation between adult IQ and, e.g., SES for identical twins raised apart. In an identical twin setup any GxE interaction would be detected since you have completely controlled for G -- both twins have exactly the same genes, but they experience different E's. All GxE effects would be part of the observed variance, which does not correlate with SES -- it turns out all to be nonshared. From your question it appears you didn't appreciate this point?
Carson: the limitation of most of the twins data (all but Turkheimer***) is that the range of environments is restricted to those families typically allowed to adopt babies, which I would characterize as working class or higher. Within that range, there is no evidence for shared environment. Whether the main effect is kids controlling/creating their own microworld, or the GxE randomization I described (which is plausibly independent of SES), or something else, I don't know.
*** Turkheimer managed to find some data on adoption into families of very low SES, and he found a much lower heritability number, although as you can tell from comments the result is controversial.
anon,
Assuming you're the same anon, and that you're attempting to respond to my point, I think you're failing to grasp my point.
There really are tests properly described primarily as achievement tests, as opposed to IQ tests. They are aptly so described because the questions themselves expect exposure to certain specific material. In contrast, tests of cognitive ability go out of their way NOT to presuppose such exposure, making sure that their questions instead only presuppose only the most basic kind of knowledge.
Thus, the Math SAT presupposes only knowledge of elementary algebra and some geometry. Effectively every HS student who takes the Math SAT should have had such exposure. For, say, the Calculus BC Achievement test, however, exposure to the material of Calculus BC class in HS is critical to a good score.
If one were to test all college bound seniors using the Calculus BC Achievement test, it's very likely that the scores would depend greatly on whether the student had taken a calculus course of the appropriate level. The correlation with g (or IQ tests) would presumably be far lower than for the SAT math test.
On the other hand, given that students who take the Calculus BC Achievement test almost always have taken the appropriate class for preparation, it's very likely that performance for that restricted range of students correlates quite well with g and IQ tests. So a high correlation with g does not by any means imply that this achievement test is in fact effectively an IQ test.
Now what of the old NMSQT? My impression of this test is that it did in fact depend much more on acquired knowledge than does, or did, the SAT. Since it was targeting the highest range of performance -- trying to identify those in the top 1% or higher -- it's likely that the amount of knowledge presupposed was considerable.
The problem is that whether a given student had sufficient grounding in that material probably depended on the quality of the schools they attended, as well as the values inculcated in the students by their families. That might certainly vary considerably among students of equal cognitive ability or g, and the degree of that variance might well depend on the SES of their families.
Now if this is true, one would expect that the NMSQT is not as heavily g loaded across all students as is, say, the SAT. And that would suggest that the correlation of the NMSQT with other IQ tests is somewhat lower than that of the SAT -- though the correlation might (and, all things considered, should) still be pretty high. I don't in any case see at first blush anything in the Turkheimer paper that addresses the issue of exactly what the correlation is between the NMSQT and IQ tests, or how well it captures g. It seems at most to point out that there's a lot of correlation between the subtests of the NMSQT (which can be explained by similar school environments as well as by g).
I think that one should take seriously Turkheimer's own cautionary point, based on the fact that the NMSQT may be regarded as an achievement test. Certainly he is aware (as you don't seem to be) that this real possibility does count against the hypothesis that heritability of g itself varies across SES, and does not just deviate at the lower end. No doubt he offers that important caveat because he knows that it is the first methodological objection that would be raised to his paper had it not been mentioned.
I did grasp that point.
IQ tests require a background, just not as much as some other tests.
When I read that the mean IQ of Hottentots is such and such, I think, "Whoever was giving an IQ test to Hottentots is a moron."
> When I read that the mean IQ of Hottentots is such and such, I think, "Whoever was giving an IQ test to Hottentots is a moron."
Possibly. I dont really know. But that doesnt really come into play when you examine schooled, literate people who are used to competing on written tests.
Ben G, Re: decline of effect of shared environment, see p. 74 of the Plomin chapter I linked to above the figure.
The studies quoted show the decline in effect by adolescence. There is no correlation between IQs of adopted and non-adopted siblings when tested in adolescence. Late adulthood is not mentioned. BTW, IQ scores before adolescence are not particularly strong predictors of adult IQ.
I suggest you study the Plomin link before writing your blog post...
The cases of "very high" IQs might help in answering to what extent IQ tests are measuring something which can't be learned.
I am almost certain vos Savant and Langan are frauds and Kim Ung-Yong is genuine, but does anyone here know?
Although with only his wiki to go by Kim is clearly a fraud also.
> The cases of "very high" IQs might help in answering to what extent IQ tests are measuring something which can't be learned.
So would a billion experiments that attempted to increase IQs.
This would show up as a correlation between adult IQ and, e.g., SES for identical twins raised apart. In an identical twin setup any GxE interaction would be detected since you have completely controlled for G -- both twins have exactly the same genes, but they experience different E's. All GxE effects would be part of the observed variance, which does not correlate with SES -- it turns out all to be nonshared.
Good point.. To be precise, such findings if true don't eliminate the possibility of shared environment by gene interactions, but rather eliminates the importance of most frequently proposed candidates (e.g. SES, parenting style).
Incidentally, Turkheimer (2007) replicated their earlier finding of shared environment mattering more for the poor, but it was unable to replicate the finding that there was an interact and IQ genes.
I haven't read much on the validity of the reared apart studies.. It seems like they are widely accepted. Then again, the fact that environmentalists like Turkheimer are out looking for interactions suggests that not everyone buys into it. If anyone has a ref which kind of summarizes the existing consensus/controvery that would be great..
As for Plomin, I'll check that out and post another comment. An important thing would be that pre-adoptive environment are taken into account, and also that there's no restriction of range per Stoolmiller (1999).
The Bajan redlegs are a great example of nurture over nature.
Heriditarians, racists, and conservatives should know of this case.
a vid on the redlegs can be seen on liveleak
What about them.
What do you guys make of Racial equality in intelligence: Predictions from a theory of
intelligence as processing by Fagan and Holland, according to which cultural differences account for the black-white IQ gap?
It's silly. How the authors distinguished general from specific info isn't explained. Perhaps the difference is easy and hard.
Re: Fagan and Holland,
The problem with research in genetics and cognition is that many (most?) people approach it with strong priors and then gladly embrace any research which confirms those priors. I suspect that your linking to this paper and Cosma hosting the pdf on his web page might be examples of this, although of course I cannot be sure.
(You could argue that I am also one of those prior-dominated people, but it's not true. For example, I was blown away by the lack of evidence for shared environment -- I've been largely convinced by the data to change my opinion of whether, e.g., SES would turn out to be an important component of shared environment in families. Before seeing the results from these studies I would have been pretty confident that such a link existed.)
As far as the paper is related to my post (and not off topic, except that the word "IQ" appears in the title), the authors want to claim that some IQ test questions are measures of specific knowledge as opposed to general information processing ability. (As anon pointed out it's not at all clear how to break things up this way.) The authors seem to think that high SES families will have had exposure to the specific knowledge, whereas low SES families would not. However if this were true then other studies would have identified family SES as a shared factor (at least within the range over which adoptions are allowed). So, that hypothesis is at least naively disfavored by the studies referenced in my post. It's certainly still possible that being Black or White is in fact a shared environmental factor, but SES within a "humane range" does not seem to be. You might not care about this distinction, but it is much easier for a policy maker to try to equalize SES than the other thing.
Hey guys, I found a very good passage on the lasting effect of c^2 in The Nurture Assumption:
"Most behavioral genetic studies probably underestimate the long-term effects of adoption because the researchers didn't make a special effort (as the French researchers did) to find adoptees reared in home that vary widely in socioeconomic status. Most of their subjects were reared in middle-class homes and middle-class neighborhoods. The dearth of adoptees reared in working-class environments makes it harder to get an accurate picture of the environmental effects on IQ.*
Fortunately, there are other kinds of data. Studies in which adoptees of various ages are compared to their biological and adoptive relatives confirm that, as they get older they become more similar in IQ to their biological parents and siblings, and less similar to their adoptive parents and siblings. However, even in adulthood they retain some advantage over biologicalo siblings who were not given up for adoption and who grew up in the family they were born into... So we can say that adoption does have a long-term effect on IQ, though it's a small one. Not twelve IQ points, but perhaps seven."**
* Stoolmiller 1999
** Loehlin, Horn, & Ernst, 2007; van IJzendoorn, Juffer, & Klein Poelhuis, 2005.
...
So you folks who say there is no lasting c^2 effect on IQ, what say you to that?
Its certainly of much interest.
Just FYI, Steve or someone did raise the SES of adopters. In this thread or the last.
Ben is this one of your/Harris' two papers, or are there multiple works in 2007 by the same set of authors:
http://www.springerlink.com/content/a4n78135lq880405/fulltext.pdf
I gave it the 25-second ADD once-over; I dont see adult IQ data.
OK, pardon me:
"In a followup about 10 years later, as adolescents or young adults,
the children from 181 of the 300 families received IQ and personality tests, and were rated again on the 24
traits by a parent (Loehlin et al. 1987)."
Where the IQ data at? Maybe its only in the Loehlin et '87 cite -- ?
OK, if I'm not mistaken Loehlin mis-cited himself.
LOEHLIN ET '87:
"The results for IQ will be the subject of a separate report."
What he meant to cite must be this:
http://www.jstor.org/pss/1131039
Loehlin et '89:
"At the time of testing, the index
children were between 3 and 14 years old. Approximately 10 years later, 258 adopted children in 181 of the families were located and retested, as were 93 biological children of
the adoptive parents."
Hmmmm. 3 + ~10 = ~13. So, am I looking at the right thing, or are there actual adult IQ data in Loehlin et 07, that I missed noticing?
Also, in Loehlin, only some of the people were relocated for the followup. This creates an additional liability, possible selection bias, which should be added to the fact that not all the subjects were apparently over 18.
In the Texas Adoption Project, some 1,230 members of 300 Texas families that had adopted one or more children from a churchrelated
home for unwed mothers were given
IQ and other tests. Persons tested included the adopted child, the adoptive parents, and other available biological or adopted children in the family. At the time of testing, the index
children were between 3 and 14 years old.
Approximately 10 years later, 258 adopted children in 181 of the families were located and retested, as were 93 biological children of the adoptive parents.
Moving on, the second of Ben's/Harris' cites is IJzendoorn et 05, a large meta-review.
http://www.apa.org.proxy.lib.fsu.edu/journals/features/bul1312301.pdf
A look at the large table 1 shows that over 85% of the papers here reviewed, assessed the IQ of adoptees of age 18 max. There are two papers examining adults only. Those would be interesting to look at.
I didnt read the text, just searched the word "adult." Thier opinion is not unorthodox, but on the other hand certainly does not rule out ANY role for c^2 in adults:
In longitudinal studies, the IQ of adopted children has been found to become more similar to the IQ of their birth parents with increasing age (Fulker, DeFries, & Plomin, 1988; Plomin et al., 1997), and in adulthood the correlation between the IQ of adopted children and that of their adoptive parents appears to be much lower than the correlation with the IQ of the biological parents (McGue, Bouchard, Iacono, & Lykken, 1993; Plomin et al., 1997).
Differences in identical twins has been attributed to differences in their epigenomes.
when the twins are discordant for cancer, autism (the real thing), ...
or so I learned from NOVA.
anon,
I think epigenetics is a bit of a buzz thing. There are possiblities and some rock-solid phenomena for sure. But I think there is still a lot of question about whether this explains much in the way of biological variation, or just a little.
Rare, mildly deleterious alleles that are present due to the mutation-selection balance, are a more logical source of most disease, and of most non-optimality of fitness in general. I cant guarantee that they cause most disease, but it seems likely. Obviously, this becomes something of an ideolog-ized issue. There is mixed evidence, for example, on whether sexual selection is partly selection for good genes resulting in higher general fitness. And there seems to be disproportionate, but understandable, emotional tinge to this debate.
If anyone wants to look up the cites on adults in the big metareview cited by Harris, here they are. I might look at em later, or not.
Smyer et al 98. Childhood
adoption: Long-term effects in adulthood.
Teasdale and Owen 86. The influence of paternal social
class on intelligence and educational level in male adoptees and nonadoptees.
Does anyone here know if vos Savant or Langan or any other purported super high IQs are genuine.
I am almost certain they are frauds, but if true these would demostrate that IQ tests measure something which can't be learned.
The high IQ societies are of course ridiculous and despicable, but still are there people from humble backgrounds who hit the absolute ceiling, score 19 on all eleven WISC subtests, for example?
anon, there is some other way of calculating IQ for children, which involves dividing by the child's age. Some claimed super-IQ scores might derive from that. This definition is never, or almost never, used or acknowledged in serious modern discussions. Whether it may have been taken seriously some decades back, I dont know.
Steve is correct in his observations. The people who have tried to argue that heritability is not really what is claimed are expressing ignorance of the issue. Heritability rises from around 30-40% in children to over 80% in mature adults. This rise is partly (significantly) boosted by the complete disappearance of the shared environment component. Those who think they can imagine errore in twin studies are apparently unaware that virtually identical results are obtained from Falconer's formula and path analysis, neither of which involve MZA studies. When three independent methodologies converge on the same number, the odds of that number being correct (especially given that all three methods are theoretically sound) are huge. But there is also a fourth way to determine heritability. It is simply to subtract the known environmental variance from 100%. The result of that exercise produces a heritability that is higher than any of the other three methods; Lynn gives 96%. [Lynn, Richard (2006). Race Differences in Intelligence: An Evolutionary Analysis, Washington Summit Publishers, Georgia.]
One way to demonstrate the vanishing nature of the shared environment is to look at non-twin siblings. When reared together, their IQ correlation is 0.49 in adulthood. When reared apart, their IQs correlate at 0.24 as children, but this rises to 0.49 in adulthood. Unrelated children, reared together (adopted) correlate at 0.25 in childhood and 0.01 in adulthood. [Rushton, J.P. and Jensen, A.R. (2005). Thirty Years of Research on Race Differences in Cognitive Ability. Psychology, Public Policy, and Law, Vol. 11, No. 2, 235–294.]
Steve wrote:
"Environmental effects cause regression to the mean of a child relative to the parental midpoint. Parents who are well above average likely benefited from a good match between their environment and individual proclivities, as well as from good genes. This match is difficult to replicate for their children -- only genes are passed on with certainty. "
I disagree with the claim that regression to the mean is caused by environmental effects; it is genetic. Murray has written about this as have Rushton and Jensen.
"Regression toward the mean provides still another method of testing if the group differences are genetic. Regression toward the mean is seen, on average, when individuals with high IQ scores mate and their children show lower scores than their parents. This is because the parents pass on some, but not all, of their genes to their offspring. The converse happens for low IQ parents; they have children with somewhat higher IQs. Although parents pass on a random half of their genes to their offspring, they cannot pass on the particular combinations of genes that cause their own exceptionality. This is analogous to rolling a pair of dice and having them come up two 6's or two 1's. The odds are that on the next roll, you will get some value that is not quite as high (or as low). Physical and psychological traits involving dominant and recessive genes show some regres-sion effect. Genetic theory predicts the magnitude of the regression effect to be smaller the closer the degree of kinship between the individuals being compared (e.g., identical twin > full-sibling or parent–child > half-sibling). Culture-only theory makes no systematic or quantitative predictions."
Rushton, J.P. and Jensen, A.R. (2005). Thirty Years of Research on Race Differences in Cognitive Ability. Psychology, Public Policy, and Law, Vol. 11, No. 2, 235-294.
Bob, why do you state that the heritability is 0.80, but also say that the first three techniques produce the "same number"? If MZ twins raised apart show a correlation of 0.49, then that is the h^2, right? But if the first three techniques all give near 0.49, why do you say 0.80?
> It is simply to subtract the known environmental variance from 100%.
Why would this work? What about epistasis, dominance, G-E interaction and correlation? Are all those, in what you are saying, included in "environment" as they often (confusingly) are?
eric,
I haven't read through the studies. I plan to do that and create a post at gnxp.com on the shared environment debate.
Some of your concerns seem uncalled for. For example, attrition exists in every major adoption study, especially in the ones most highly cited by hereditarians (e.g. the MTRA).
BobW,
Your comments on shared environment ignore the restriction of range issue. Studies which make sure to sample more from the impovershed reveal a significant (but not huge) shared environmental effect.
Also, regression to the mean makes sense from an environmental perspective. If people succeed both because of their genes AND their environment, we should expect that people that have exceptional genes AND exceptional environments relative to their populations will not strike such a lucky combination with their kids most of the time.
I believe my description of regression is correct if one assumes additivity of genetic influences. (Then the trait value to which the children regress is the parental midpoint.) But interactions could also produce some additional regression, as described in Bob's excerpt which cites "combinations" of genes as important. I always thought that additivity was a decent first approximation, but would like to know if more is known about this from actual studies of IQ.
In any case the assumption of a random environmental component (uncorrelated between generations, which is consistent with dominance of nonshared effects) leads to regression all by itself, and even for large heritabilities like .8 I would be surprised if the effect from non-additivity at the genetic level gives a larger effect that the environmental component. To be honest I think Rushton and Jensen are just plain confused in that excerpt, but perhaps I am missing something.
See, e.g., below for more discussion (standard textbook on population genetics):
http://books.google.com/books?id=xf-ROozfS14C&lpg=PA200&dq=population%20genetics%20regression%20mean&client=safari&pg=PA140#v=onepage&q=&f=true
Ben, In a drug trial it is obvious a priori that side effects and weakness of the desired effect are likely to affect who drops out of the study. Here, I admit theres not a strong bias in who drops out thats obvious a priori, but who knows? Even weak effects are salient here because we are looking for a change in IQ that we know will be weak at best.
"it is genetic" says Bob W.
Regrssion to the mean is expected whatever the heritability.
Rushton, Jensen, and Lynn would explain the Afrikaner British IQ gap on genes or racial admixture and the Flynn effect on nutrition.
When someone quotes these people that someone should be ignored.
Yer wrong. Jensen in particular is well regarded even by non-fans.
I suppose your point about the Bajan Redlegs is that they aint as rich or educated as Americans and probably have lower IQ. Well, the people that were moved there were according to what I read "the poor, the hungry", and recalcitrant landlords Catholic clergy. I'm guessing the recalcitrants were the least numerous.
The exact quote: "In the meantime what we know is that Cromwell decreed that troublemakers – the poor, the hungry, clergy and Catholic landlords who refused to move to Connacht – be sent to Barbados."
Wikipedia states "Since independence from Britain in 1966 when most white Barbadians left for United Kingdom,[...]" -- so theres a *second* selection event that left the poorest and the least bright in the Redlegs population. If you think their relative poverty or ignorance cries out to be explained by environment or history, I dont agree.
I dont know much about Afrikaaners but the same may well apply.
Whether other peoples with enslavement history have ever been selected, our double-selected in the way of the Redlegs, I couldnt say.
The Aussies come from even worse stock!
Well, this is like arguing over transsubstantiation with a Hindu. Pointless.
But I'll try one more example.
Obesity is as heritable as IQ in some poulations, yet there is no reason it needs to exist at all.
This should deconstruct the whole debate.
The level of science is so low in this "field" it can't be taken seriously.
To those who say no intervention works I say neither do diets.
Bob W and Jensen could find a g for athletic accomplishment after taking measurements, tissue samples, etc.
Then Bob W would proclaim that height has nothing to do with ability at basketball.
Yeah, I suggested that a low IQ subset of Irish were moved to Barbados, and later a higher-IQ subset of that subset moved back to the British Isles. Definitely a mind-blowing supernatural event -- bow down and worship if you fear God.
I wouldnt say my idea about the Redlegs definitely explains everthing with 100% confidence. All I said is that its far, very far, from a screamingly obvious case of whatever anti-hereditarian phenomenon you imagine to be at work.
> Then Bob W would proclaim that height has nothing to do with ability at basketball.
No one has claimed that there is a single task in the world that loads solely on /g/. You know that. You know I know you know that. Arguably, your posts on this thread are a great example of a cognitive task that is not particularly g-loaded.
You should understand what I've written before you comment on it.
And by understand I don't mean agree.
the redlegs were just as impoverished before independence.
oh well you say things were different then. then environment mattered.
your take on the redlegs was stupid.
Bob W claimed that no IQ factor other than g was related to job performance. Dum, du, du, du, dumb!
Eric,
Lynn’s average MZA h^2 = 0.75. Corrected for attenuation, this becomes h^2 = 0.83. This number is consistent with heritability findings for specific studies published in the past year in Intelligence.
Steve wrote: "In any case the assumption of a random environmental component (uncorrelated between generations, which is consistent with dominance of nonshared effects) leads to regression all by itself, and even for large heritabilities like .8 I would be surprised if the effect from non-additivity at the genetic level gives a larger effect that the environmental component. To be honest I think Rushton and Jensen are just plain confused in that excerpt, but perhaps I am missing something."
Regression to the mean goes both up and down (statistically increases IQ when it is below the mean). All of the environmental factors that have been identified (please correct me with examples, if any exist) are detrimental to IQ. How do you see increased IQ resulting from environmental effects? What studies have demonstrated this?
In Lynn, R. (2008). The Global Bell Curve: Race, IQ, and Inequality Worldwide. Augusta, GA: Washington Summit Publishers, there is a discussion of how regression to the mean probably worked in causing East Asian IQs to rise in Brazil and Africa.
related... Breastfeeding is not an IQ boosting candidate, since it is the default. Formula feeding may cause lower IQ by the same process as is seen when there are deficiencies in iron, iodine, and folate. The magnitude of the difference between formula feeding was grossly overstated in 2007 by Moffitt and Caspi because they refused to correct their data for maternal IQ. This was not an oversight, but a refusal. I discussed the paper with Ian Deary, who was on the jury for the paper; he said that he pointed this deficiency out and asked the authors to make the correction. Even though they had the necessary data, they did not make the correction and elected to overstate the IQ differences.
"Regression to the mean goes both up and down (statistically increases IQ when it is below the mean). All of the environmental factors that have been identified (please correct me with examples, if any exist) are detrimental to IQ. How do you see increased IQ resulting from environmental effects? What studies have demonstrated this?"
OK, I didn't appreciate that this is the point you were making. I personally find that to be a radical view, but since we know very little about nonshared environment (within which one would find any positive factors for IQ development), I agree that there are no specific studies or factors that I can point to.
Just to clarify for others -- the model I used for regression is simply X = G + E, where X is the exhibited phenotype, G the genotype and E the (uncorrelated random variable) environmental factor, which can be both positive and negative. Bob is arguing that we only know of negative E effects, but not positive ones.
However, the sign of E is not well defined as I can shift both G and E by a constant leaving X unchanged and make an E which originally had both positive and negative signs into one which is negative definite. (This assumes the range of E values is finite, which is not unreasonable.) Thus, we really only need to refer to environments as less good than the ideal (max E) environment.
**Probably the model you have in mind is that max E is achieved in most cases and lower values of E appear only due to some exceptional deprivation -- the issue is not the sign of E but rather the shape of the distribution** (The "standard model" -- see population genetics link in previous comment -- is just to take E to be a Gaussian random variable. As usual, since the effect depends on many variables, there are Central Limit Thm reasons to posit this as a first guess.)
Note that to achieve the regression effect you mentioned -- below average parents tend to have less below average kids -- we need only invoke a negative environmental effect. Parents who are well below average are likely to have had a more negative than average E, and the kids (again assuming E a random variable) are likely to have a less negative E than the parents, leading to a higher X. It's to explain the regression in above average families that the positive E values in the original model are necessary -- the parents were lucky in E and G, the kids get the same (parental midpoint, assuming additivity) G but likely a smaller E, so kids regress and are less above average. Note, using the "shift symmetry" both of these effect persist in a negative definite E model.
Please correct me if I am confused.
BTW, since we do observe regression in above average families, I assume you want to attribute this to non-additive genetic effects? I thought the evidence pointed to small non-additivity? Also, I think in your model regression of below average families is very different than for above average families, since E is not symmetrical. Is there any data on this?
To summarize my long comment:
1. Bob thinks max E (best environment possible) is often realized, whereas I suspect E has a broad probability distribution and max E is not often realized.
It is easy to detect E's which far below average, due to severe deprivation. But, since almost all of E is nonshared in the range of "humane environments", it is not easy to quantitatively measure the whole range of possible E's.
2. I suspect genetic effects are to good approximation additive, but Bob's model depends crucially on non-additivity.
3. In the case of human height, and in various animal breeding experiments (where the formalism has been widely applied), I think the standard Gaussian model for E works pretty well. But I would welcome pointers to data.
Steve --- “Note that to achieve the regression effect you mentioned -- below average parents tend to have less below average kids -- we need only invoke a negative environmental effect.”
Maybe I should have copied in all of the 30 year review comments on this. The reason for the discussion there was to show that regression to the mean goes to the group mean. Blacks regress to 85 and whites to 100. This happens even when the data is limited to a single narrow SES range. How would environmental effects cause such a large and consistent regression, especially when they act only in one direction? If you have not read the Rushton/Jensen paper, I commend it as excellent. You can easily locate it with a search engine. The discussion of regression to the mean was presented in such a way (different racial groups) that it can only be explained by genetic differences. FWIW, I do not believe that Murray, Rushton, and Jensen are all confused over this phenomenon.
In the X = G + E model the regression in a *particular* family is to parental midpoint G, but if you average over a large population the regression will be to the *average* parental midpoint G, which would be the group mean, or average parental midpoint within a population. In that sense you get regression to group mean.
(You can't test these models on a single family, as the number of kids is too small, the data is too noisy, etc. But of course the *model* does describe individual families. When we want to compare to data we are forced to average over large groups.)
I don't think there is necessarily any contradiction between the two models on the point that Rushton and Jensen are trying to make.
But I do think what they wrote in the excerpt you posted is confused. At least, it's not the usual model of regression used by population geneticists (including, for example, people who really do things like breed livestock and plants).
If you know those guys, please forward this discussion to them as I would like to hear their response. Again, I'm not disputing the use of regression to investigate the group difference question they are interested in. Even in the "standard" model the regression, when averaged over a population, is to the group mean. (I haven't checked this carefully, but I think it's correct.)
An environmentalist could point out that racial groups are associated with cultures. For this reason, regression to the mean of ones racial group need not be a genetic occurrence. I went to a mostly black school where the parents of black kids were doctors/lawyers (this was one of the richest black communities) but their kids underperformed. An environmentalist could say that the kids underperformed because they took on the cultural patterns typical of their group. Whether this happens because of genetics or because of environmental factors is the very crux of the debate. So I don't think regression to the mean is a hereditarian ace in the hole.
Ben G,
No offense to your "g" but I think you are missing the point.
The post is about the difficulty in establishing the existence of any form of shared E.
In your comment you just postulated an "ethnic" shared E. While such a thing might exist it is strange that one can't detect, e.g., SES as a shared E whereas you are confident about the existence of the one you just postulated.
Before seeing the data I was confident about the existence of an SES shared E (as were all egalitarians), but it turns out I was wrong. Isn't it possible you are wrong now?
Steve,
Systematic differences in nonshared environments could also create gaps between groups. That is Harris's view regarding the gaps between blacks and whites.
I'm interested in your take on the excerpt I quoted, by the way.
Ben G,
"Systematic differences in nonshared environments" would constitute, in and of themselves, a shared environment. Think about it.
Instead of the old claim "SES systematically affects individual learning environments" you are substituting "ethnic group systematically affects individual learning environments". If the former (to all egalitarians' surprise; I was surprised too) turned out to be untrue (or too small an effect to explain group differences), then why do you think the second hypothesis will turn out to be true? Because you don't like the alternative conclusion?
Do you realize that you sound like a prior-driven confirmation bias machine rather than a learning algorithm?
Systematic differences must be very closely controlled, if they have anything to do with the B_W gap. Blacks fall at the bottom of the intelligence spectrum worldwide, including in countries that are 99% black and in countries that are governed by blacks. In every instance where blacks and East Asians are found in the same country, the East Asians lie at the highest SES level, with blacks at the lowest. This includes sub-Saharan Africa, Brazil, Europe, and the US.
BobW,
I looked at the paper you referenced:
http://www.udel.edu/educ/gottfredson/30years/Rushton-Jensen30years.pdf
I think p.263 is the relevant one.
Paragraph 1 (P1) looks like the one you quoted. It suggests that they think regression is due to non-additivity of genetic effects, which is not, I believe, the conventional view. (Non-additivity might contribute, but random environmental effects are probably larger, even if heritability is as high as .8 .)
P2, P3: If the data is as they describe it supports their conclusion, at least qualitatively, even in the X = G + E description of regression. (See my earlier comments.)
P2: "High IQ Black parents do not pass on the full measure of their genetic advantage to their children, even though they gave them a good upbringing and good schools, often better than their own. (The same, of course, applies to high IQ White parents.) Culture-only theory cannot predict these results but must argue that cultural factors somehow imitate the effect theoretically predicted by genetic theory, which have also been demonstrated in studies of physical traits and in animals. "
When they refer to "genetic theory" as something which is tested in animal studies, I believe they are referring to the standard model for regression to the mean, which involves both G and E (see link in earlier comment to textbook treatment).
My conclusion is that they have a nonstandard view of regression -- I guess psychometricians are not population geneticists :-( Given that they don't comment on this, I have to think they don't understand the standard view. (The standard view would lead to similar conclusions about group differences, but not rely on non-additivity as the main effect. Perhaps they aren't aware of nonshared dominance, hence are unwilling to consider a random environmental effect?)
Note that historically even Galton was confused about regression to the mean and it took Pearson to straighten him out (the latter a Wrangler, the former not ;-)
Steve,
I don't actually believe in either
the hereditarian hypothesis or the environmental hypothesis, because I'm still learning at this point. That's why I'm doing so much reading and asking so many questions..
On the subject of shared environment: I think a first point to address is that excerpt. If true, then the shared environment (likely through the socioeconomic status of one's family) does have a lasting impact.
As for this point:
"Systematic differences in nonshared environments" would constitute, in and of themselves, a shared environment.
I didn't think of that.. I'm gonna hit the books for a bit to get a better understanding of how group environmental differences could plausibly play out..
1. That some shared environments have little ot no effect does not mean that all shared environments have little or no effect.
2. The appropriateness of the bell curve model for environmental effects depends on the trait.
If, within a population, IQ were like height where most today reach their genetic potential (No?) then the bell curve model is not appropriate.
3. Still the example of obesity shows this entire debate to be silly.
a. obesity is as heritable as IQ
b. obesity rates vary with ethnicity
c. diets don't work
d. there is no reason why anyone must be obese
I see Bob is on again about the IQ of sub-Saharan Africans.
The obesity example shows two things:
1. Modify the range of environments and you'll alter the range of phenotypes and the heritability.
2. Heritability is not equivalent to malleability. And failed interventions don't change that fact.
I don't think any of the hereditarians deny either of these points.. So I don't see how the IQ debate is silly..
As an adopted child, my upbringing benefited because my (adoptive) parents didn't share my own weaknesses (e.g., procrastination). With my own children, their biological tendencies toward procrastination are exacerbated by a pro-procrastination upbringing.
I would add that siblings sometimes choose to be more different from each other than they would be if raised apart simply to assert a distinctive identity from their siblings.
Regarding restriction of range in adoption studies:
The French adoption study with a tiny sample size (n = 38) that looked for children adopted across class lines found that being born to the highest class added 16 points to IQ and being raised in the highest class added 12 points to IQ (measured at age 14).
That seems pretty plausible to me: that both nature and nurture matter in IQ, with nature a little more important.
Identical twins sometimes wind up in different environments because they were raised together. For example, the identical twin basketball players Horace and Harvey Grant were raised together, so Harvey played shooting forward and Horace played rebounding forward. Horace Grant became one of the skinniest power forwards in the NBA. If he had been raised separately from this twin, he probably would have grown up to be a quick forward.
I asked Turkheimer what the mean IQ was in his famous study, but he wouldn't tell me.
Think about the obesity example more.
twin studies could be taken more seriously if they were for twins seprated by country. has this been done?
one twin goes to trondheim the other to Durban. one to Tokyo, the other to Bangor.
The French adoption study with a tiny sample size (n = 38) that looked for children adopted across class lines found that being born to the highest class added 16 points to IQ and being raised in the highest class added 12 points to IQ (measured at age 14).
Maybe the 4 pts can be explained by the low class mothers or infants not getting enough good food to eat, breathing in crappier air, or some similar prenatal/infant environmental factor.
How do you factor out these types of environmental differences anyway? The lower classes live differently than the upper classes. This is true even when the lower class person is pregnant.
I'm open to the idea that restriction of range masks the existence of shared environment (e.g., SES correlations) over a broader range. That's why I keep citing Turkheimer's adoption result (reduced h) as an interesting possibility.
The French study is interesting, but the size of the data set (n=38) is way too small. Compare to the stuff Plomin cites (e.g., Colorado adoption study with 1000+ kids), or his ongoing work which has access to all UK adoptions in a 1-2 year period.
But let me emphasize something about the post (again): it's not mainly about the group differences (race) issue. It was actually motivated by the parenting issue: how much can a middle or upper class parent really expect to enhance their kids' environment? If the data doesn't show any SES benefit (e.g., to adopted kids) across middle and upper class backgrounds, it suggests something about how hard it is to systematically enrich your kids' learning. Forget about the B-W thing, it's an interesting observation even within a particular ethnic group.
From Steve Sailer's comments it seems he finds my learning style and environment matching factors to be plausible. I do as well, and it suggests that a lot of what affects cognitive development of kids is out of the parents' and the state's control (especially within the middle to upper class range). Also, I think Steve understands the point that max E is rarely achieved -- lots of things have to go right for someone to achieve their maximum potential. It's not just a matter of deprivation, things have to go right as well. (The right teacher, coach, mentor, best friend, book, etc. can have a big impact.)
Re: obesity, that's an interesting point, but it's seems to me that bodyweight is much more malleable under environment influence. You can't change your IQ very much after adolescence, but of course your weight can change radically over a year or two at any point in your life. There is a precarious positive/negative calorie balance that is strongly affected by environment. It seems qualitatively different to me than IQ. The point that this is an example where heritability measured in a certain way can be very misleading is well taken, but we've been spending billions to enhance IQ (test scores) for decades now with little result, so it's not as if we know nothing about response of IQ to environment. So far I've seen no evidence for *systematic* methods of improving cognitive development, at least beyond what is already available to, e.g., working or middle class kids.
anon, The obesity example just shows that the mix of native and external influences can change over time, or differ for different pops. We know that. I assume you would see a robustly larger environmental component for adult IQ in subsaharan Africa. Likewise, if we all started being hungry often here in the West including in childhood, and men started growing to 5'4" on average, then the heritability of IQ and height would decline here.
Are there any transnational twin studies?
If there are, you should look at countries with similar wealth. A lot of US foreign adoptees are from China, I think. Though it wont remain so for more than another 14 years tops, there are still many places there where you wouldnt want to be born. Lot of tuberculosis. Some children might be hungry enough to lower their adult height and IQ, I am not certain though. In India definitely, and that will take longer to be corrected.
from Steve's post:
>Technical remark: if n is large,
>and factors uncorrelated,
>the observed environmental
>variation in a population will be
>suppressed as n^{-1/2} relative to
>the maximum environmental effect.
That can't be right, because "n" is a modeling artefact. N could even be infinite, under some finance-flavored Brownian motion model where the child's interaction with the environment moves his "portfolio value" up or down; but this high dimensionality would not necessarily imply ginormous ratios between the maximum possible effect size and the typical effect size from the environment.
I understand your allusion to sums of correlated random effects being larger than the uncorrelated sums (i.e. the sqrt(n) factor in the Central Limit theorem). Certainly things can be far different for a kid if lots of things all go right simultaneously instead of being left for chance to sort out (win some, lose some). But as the above remarks indicate, something has to go wrong with your "N" story. Possibly as you consider finer-grained models (higher N), the different variables will have vastly different sizes of effect; there will be some hierarchy of effects and sub-effects and sub-sub-effects (epicycles). If so, maybe the parental problem is more tractable and you only need to get enough of the big components to all point in the same direction.
Good luck with the parental vector engineering.
Igon,
You are correct, as n gets larger the contribution to the overall variance from each factor goes down. In my crude model I was thinking that n had a fixed value that characterizes different distinguishable factors of a certain size. But I just wanted to make the point that n could be pretty big -- it's not just a few factors.
Steve,
I don't know if you're still reading the comments here, but I was wondering what your explanation is for adult identical twins reared together being more similar than adult identical twins reared apart. Bouchard et al (1990) found the correlation was .88 for reared-together MZ twins but only .69 for reared-apart pairs. A similar estimate for reared-aparts is available in the recent article by Bouchard (2009).
I'm not familiar with that result -- what age are they when tested? I seem to recall reading the claim that there was no significant difference.
They are adults. Here is an excerpt and list of the studies from the section "The heritability of IQ in adulthood" in Bouchard (2009):
"Table II shows the results of the five extant studies of MZA twins, who are mostly adults (Newman et al. 1937; Shields 1962; Juel-Nielsen 1980; Pedersen et al. 1992; Johnson et al. 2007). There is very little variability from study to study and they all yield estimates fairly close to that yielded by studies of ordinary twins."*
Study-----Correlation---Sample (pairs)
USA 1937--------0.71----19
Denmark 1965----0.69----12
England 1962----0.75----37
USA 2007--------0.73----74
Sweden 1992-----0.78----45
Weighted average0.74----187
*He provides no citation for that, but Bouchard (1990) implies a significant difference.
"Ordinary" twins? I guess he means MZ together but he should say that.
Just double checked Bouchard (1990). It is adults as well:
"As illustrated in table 4, however, adult MZ twins are about equally similar on most physiological and psychological traits, regardless of rearing status."
Table 4:
...
Variable--------MZA ----- MZT
...
WAIS IQ---------.69-------.88
Ben,
Those numbers are a bit surprising to me -- they seem to indicate a small shared environment for identical twins. In the Plomin links I provide he doesn't seem to acknowledge that interpretation, although I should look again.
I have read the claim in various places that twin results tend to overstate the size of shared environment, at least relative to what ordinary siblings would experience, because the twins are exactly the same age, look the same, etc. Perhaps they say this because of evidence that MZ twins do experience some shared environment. Then the more correct statement would be that ordinary sibs do not experience a shared environment.
Steve,
I find the numbers surprising too. JR Harris gave me the references in a recent reply to one of my emails. She uses this argument in No Two Alike (which I'm yet to read) as well, to show that there is a small shared environmental influence.
What I don't get (and what I just emailed her back), is how within both types of adult twin studies, shared environment doesn't show up, but between them it does. You're the math wiz here, so maybe you could explain that.
Also, on the subject of restriction of range, my recent reading supports the interpretation that it's probably less important than Stoolmiller or Harris or I have suggested. McGue (2007) explicitly tested the impact of restriction of range and it only made the adopted child-parent correlation go from .19 to .20. I also just emailed Harris about this, awaiting a response.
Ben /g/,
> WAIS IQ-----.69-----.88
One question is, how /g/-loaded is that WAIS? Some of the other traits it loads on might be more environmentally-determined than /g/.
Eric, you might be onto something:
the MZAs (.78) and MZTs (.76) scored basically the same on a Raven's Matrices - Mill-Hill Vocabulary test composite.
Raven's has the highest g-loading of any test, but I don't know about the mill-hill. my impression is that a vocab test would be less g-loaded than the WAIS, which was created to measure reasoning ability.
information processing/speed of response shows much greater similarity in MZTs (.73) than MZAs (.56). don't know how g-loaded it is..
btw you can view the table by scrolling down to pg 142 here.
Who gave an average of heritabilities? Someone who didn't know what it was.
What was the mean distance between the twins in miles?
Raven's up by 21 points in 30 years in the Netherlands.
Y'all is jus' too stupid to think about this here.
An 11 subtest battery, the WAIS, is less g loaded than the homogeneous Raven's?
This alone proves g is BS.
All y'all ain't never gonna get it is you?
An entire pop does have a (single) heritability, whether or not there be subsets with different heritabilities.
> An 11 subtest battery, the WAIS, is less g loaded than the homogeneous Raven's?
Thats right. Are you surprised that, since not all tasks are equally g-loaded, a test with a bunch of tasks is less g-loaded? You wouldnt be, if you thought straight about it for above 5 seconds.
I cant say I have the answer to the Flynn effect. But it could be that healthier mothers have better babies, and that mothers' health may depend on what they themselves experienced as babies/fetuses (which has been claimed for many diseases). It may also be that pregnant women did not eat truly repletely until food became trivially cheap after the big war. These ideas do not have high probability a priori of being true, but they are testable for both IQ and height. One day there will probably be an answer.
Better hygiene and housing might also help IQ by reducing infections of babies, as might secular changes in breast feeding. The impact of breast feeding on IQ is currently not completely clear (I think), though I havent read the papers or anything and I might have an opinion if I did.
How far back do high quality IQ data go? Bottle feeding rose in the 40s and started siking in the 70s.
There is a hypothesis that many infections declined due to increased hygiene, better eating, and other effects of wealth that are not directly of a medical nature. This is probably false; the one big infection I can find that did oddly decline a lot without an obvious medical explanation is TB. However, it could be that hygeine and wealth did, for some reason, cause a decline in infections of babies even though it did not in general. Or, improved hygiene and eating may have caused a decline in virulence of some infections, without causing a decline in cases. There are good theoretical reasons to expect this.
These are just some of the things you could think of if you knew biology.
An 11 subtest battery, the WAIS, is less g loaded than the homogeneous Raven's?
Thats right.
= CRAP
The Flynn effect, infection, healthier mothers, blah, blah, blah
= CRAP
I would advise straight thinking but given that you can't think straight or otherwise I won't.
You're carrying around a g load in your pants.
"Thats right. Are you surprised that, since not all tasks are equally g-loaded, a test with a bunch of tasks is less g-loaded?"
You haven't got a clue.
>An 11 subtest battery, the WAIS, is >less g loaded than the homogeneous >Raven's?
>
>This alone proves g is BS.
That's a good point that deserves further analysis. e.g., is the Ravens really one-dimensional (all questions count equally)? Are these tests trying to measure the same "g" (the first principal component of the same reference data set), or they are projecting onto WAIS-g and Ravens-g (1-dimensional subspaces) derived from a different reference population for each test?
Yeah, I guess that is a fair criticism.
Post a Comment