At my post yesterday on race, several comments accused me of cowardice for not taking a position there on race-IQ correlations, even though that is not directly relevant to the post, and even though I have commented on that topic before
"So why do you accept the conclusion if the argument is incomplete?"
Your argument is absolutist in tone though, by your own logic we shouldn't accept evolution per se due to the fact the chain of transitional fossils has not been 'completed' yet.
Nice to you see smugly attacking Catholics and Homeopathy practitioners on your blog though. Isn't it funny how egalitarians always pick the easiest targets to beat over the head? Not a single one would ever challenge the real dominant paradigms that rule our age, the real taboos of racial egalitarianism and how any inequality of outcome is invariably down to some kind of 'phantom discrimination' and can be corrected with social engineering.
Far from it, you support these paradigms and gleefully engage in the kind of character assassinations that followed Watson's comments years ago, or Lahn's research into Microcephalin and so on.
Fuck Zionism by the way. If you want to talk about 'religious evil' just read the damned Talmud, I've never seen so much hateful supremacist filth in all my life, the Babylonian Talmud is like Mein Kampf on steroids.
There is cross-talk between twin fetuses in utero. In cattle, there is what is known as a “free martin”, the female twin of a male calf. A free martin is usually always sterile.
The increased autism in male-male DZT is probably due to higher levels of testosterone.
Also, I only have access to the abstract for the ncbi paper. It only references the concordances for MZ and DZ twins (by gender). Could you give the concordance for full-siblings here?
That's pretty neat, Jason. I would have assumed that the variance in Mendelian segregation, as well as the portion of the genome affecting height, was small enough that it would be hard to find anything that way.
Like the commenter "Miko" there, I am outside my domain competence. But in browsing the paper they only reference gender & age as assumed to have linear effects. Also, daedaulus2u, your plos link is broken.
Speaking of pre-natal effects, I have occasionally heard (such as from the second video here) that crack babies don't actually turn out worse than comparable children not exposed to crack. That's surprising to me, do any of you know much about it?
Uh Jason, you should look at the comment in the PloS article you linked to. It is not “assumption-free”, it is “fewer assumptions”. As was ably pointed out, they still assume that gene effects are not interactive other than being linearly additive. We know that genes have effects that are not linearly additive. That analysis assumes that those interactive effects are all negligible. We know they are not.
If you look at a gene like MeCP2, we now know that loss of it causes Rett Syndrome in females and is fatal in males (pretty strong rescue of the phenotype by interaction with an X chromosome which happens to be inactive in the cells that are rescued). The MeCP2 protein binds to methylated DNA, and so MeCP2 regulates how DNA that has been epigenetically programmed by DNA methylation is read out. MeCP2 doesn't have any direct effects on phenotype, the only effects are secondary, mediated through differential effects on the transcription of other genes.
DNA methyl transferases also only have effects through differential regulation of other genes. SNPs of DNA methyl transferases do seem to have effects on intelligence.
Presumably those DNA methyl transferases are having effects through their normal function, that of methylating DNA and so affecting how it is transcribed.
How about genes for transcription factors? They don't have direct effects either, only effects mediated through differential transcription of other genes. Something like 5% of the human genome is transcription factors. They only have effects mediated through gene-gene interactions.
People have actually looked for the genes that supposedly cause intelligence to be inherited. Have they been found yet? No, they have not. Until the specific genes are found and reliably identified as influencing intelligence and under what environmental conditions, their existence remains hypothetical. Due to the level of effort put into finding such genes, and the essentially complete lack of success in finding them, to me their existence is becoming increasingly speculative.
Pretending that models of inheritance that ignore gene-gene interactions have risen to the level of “fact” is disingenuous, or as you would say, “a lie”.
Uh, we know a lot more than was known in 1950. We now know that the actual genetic material is DNA, and we now know a lot about how it is replicated, stored, regulated, epigenetically programmed, silenced and transcribed. We have actual data on similarities of twins vs siblings, and fraternal twins. We now know that fraternal twins have a higher concordance for autism than do full siblings.
Sharing an in utero environment does make fraternal twins more similar along the autism spectrum than full siblings. If fraternal twins have a higher incidence of autism, then presumably there is something about their shared in utero environment that affects the normal process of neurodevelopment and causes autism. Presumably there are also sub-clinical effects of what ever that in utero environmental process is that produce effects less than autism. Presumably those effects might include something like intelligence.
TGGP: "we can distinguish between identical twins, fraternal twins, and unrelated children raised together (or apart). This is how measures of heritability are obtained"
daedalus2u: "The major “data” linking genes to intelligence is from a naïve examination of monozygous twins. Monozygous twins share a lot more than just a genome, they also shared pretty much the same in utero environment where their brains grew from a single cell to 100,000,000,000 cells, a factor of 10^11. "
Evidence for the high heritability of intelligence, and other traits, is extensive, multifaceted, and unimpeachable. It is not based on one method, but on numerous, mutually reinforcing methods. Whatever working assumptions behavior genetic methods may have started with, have long since entered the realm of supported fact.
Genetic denialism is quackier than ever. We don't even need different levels of kinship anymore (e.g. identical twins, cousins, etc), we can now demonstrate substantial heritability using nothing but full siblings! Welcome to the Post-Genomic Era; take this flyer and please remember to get your hand stamped at the door.
And not that it matters (since the assertion it was tied to was a lie), but sharing a prenatal environment overwhelmingly makes siblings less alike, not more alike.
Yes, when I said "imperfectly measured" that was a reference to the flawed nature of the procedure. I think you now exaggerates its problems though. As Robyn Dawes explained in "Rational Choice in an Uncertain World", statistical models with randomly chosen coefficients do surprisingly well. The important thing is that they have the right sign.
"The HH requires modeling intelligence as a linear combination of genes and environmental influences"The hereditarian hypothesis is that a substantial portion of the variation in intelligence is due to genetic variation (or alternatively, that the same is true for variation among population groups). If IQ was attributable purely to a massively complex interaction of countless genes, hereditarianism would be obviously correct. If IQ was purely the result of a simple linear combination of environmental factors, hereditarianism would be obviously incorrect.
"A genotype growing up in one environment may have a high IQ phenotype and in another environment may have a low IQ phenotype. One environment may produce a high IQ in one genotype and a low IQ in another."That can be the case even in a linear interaction model. You could correct it by specifying that in the first sentence you are speaking relative to a different genotype but the same environment, and vice versa for the second sentence. Unfortunately, we don't yet know enough about genetics to make such comparisons. We can observe that there are still large amounts of variation for identical twins with very similar environments, but that's compatible with a linear interaction model in which random noise plays a large role.
On stress and measurement: we can distinguish between identical twins, fraternal twins, and unrelated children raised together (or apart). This is how measures of heritability are obtained (though they can only directly shed light on contributions to within-group variation rather than between group variation, biracial children perhaps shedding light on the latter). We don't have equivalent comparable classes for stress.
But if there are ten-dollar bills on the sidewalk researchers have overlooked in the form of unmeasured outcomes, perhaps you should start contacting researchers and offering suggestions. I often email professors out of the blue and they can be surprisingly receptive. There is a subfield of psychology devoted to childhood development, and I would expect that is your best bet.
You do appreciate that the procedure I outlined for calculating height from bone length is flawed, even though height is (essentially) a linear combination of the lengths of certain bones? The coefficients are pretty close to either zero or one. Zero for bones like ribs and 1 for bones like the femur. If a calculating procedure like that fails for something as simple as height and bone length, why would anyone expect such a procedure to work for something as complex and as poorly defined as intelligence? It would virtually certainly work less well.
Intelligence is not a linear combination of anything. It is highly non-linear. The problem is that the interactions between the genes and the environment are non-linear and the interactions are coupled.
The HH requires modeling intelligence as a linear combination of genes and environmental influences of the form (where G1 is gene 1 and EI2 is environmental influence 2 and g1 is the constant that determines the effect of G1):
In reality, the g's and e's interact with each other and with themselves. The first coefficient isn't a constant, it is a non-linear function of all the G's and all the E's and all their mutual interactions that collectively influence the effect of G1 on intelligence.
It is a non-linear function with at least hundreds of genetic terms and likely thousands of environmental terms. These functions are different for each different tissue compartment and cell type that has an influence on neurodevelopment and brain function, heart; liver, kidneys, immune system and vary over time and exhibit hysteresis. It is also epigenetically programmed in utero and maybe in parental gametes.
I appreciate that the actual complexity of the genetic influence on intelligence is too complex to model given our current understanding of genetics and environmental influences. The HH requires neglecting all this complexity and ignoring all gene-gene interactions, all environment-environment interactions and all gene-environment interactions. We know that is not correct.
Let me restate it a different way; it is extremely likely that genes that affects IQ can have both positive and negative effects on IQ depending on other genes and on the environment. The effects of a particular gene on IQ are not just “small”, they can be either positive or negative depending on other things. A genotype growing up in one environment may have a high IQ phenotype and in another environment may have a low IQ phenotype. One environment may produce a high IQ in one genotype and a low IQ in another. Environments that are nominally “the same” can produce completely different results even in the same genome, as in the case of MZT being discordant for anencephalopathy.
Many of the genes that affect IQ are active in utero, when the brain grows from a single cell to 10^11 cells. Epigenetic programming of DNA in utero is known to affect animal behaviors as adults. Exposure to stress in utero is known to affect adult behaviors in animals and in humans. Are there differential levels of stress between the B/W population? Are pregnant white women exposed to more or less stress than pregnant black women? What is the basis for ignoring those differential levels of stress? Because stress levels are hard to measure? Gene frequency hasn't been measured at all, but the HH has no difficulty attributing large effects to unmeasured and unknown genes while completely neglecting in utero stress which is known to be different, and could be measured. Certainly outcomes attributable to stressful pregnancy are known (low birth weight) and could be measured but are not.
Yes, height can be thought of as the sum of many smaller components, but it is still a unitary trait: it can be accurately expressed with a single scalar. In your analogy there is a trait which is real and unitary in nature but imperfectly measured.
I have never heard anyone say that IQ is the result of a single "IQ gene". The standard take among hereditarians I've heard is that there are a very large number of genes with individually small effects. The addition of such large numbers of small factors results in normal distributions. There are also of course some genes with large effects that give rise to retardation.
Height is not a unitary measure of any physiological primitive. Height is the sum of the lengths of the femur, the tibia, the individual bones in the spine, the diameter of the skull, the thickness of the scalp, the hip, the foot bones, the thickness of the foot pad, plus the thickness of cartilage between each pair of bones.
The way that the length of a bone like the femur is determined, is primarily when the growth plates stop growing, stop adding length to the ends of the bone. That growth is determined locally, by the cells that comprise the growth plate. Skull diameter is mostly determined by brain size, but skull shape matters too. Should we expect there to be a single unitary “height gene” that synchronously controls the proportionate size of the height component of each bone independent of other physiological parameters? No, we shouldn't.
Suppose you didn't have the ability to measure distances longer than 1 meter, but still wanted to find the “true” or “quintessential height” (QH) of an individual by taking a weighted sum of the lengths of the individual bones. Since you are not quite sure how the different bone lengths will add up, or which bones will be important, you decide to take a weighted sum of all the bones you can measure and determine which ones are most important.
Assuming you can measure 200 bones, you measure those individual bone lengths of 100 individuals and then determine the weighting factors to use in weighting the individual bone lengths to quintessential height. You don't know what quintessential height actually is, so you use the weighting factors that give you the “best” fit but don't know exactly what weight to put to each of them, so you solve for the weighting factors that will give you the “best” overall fit. You now have equations of the form
You still don't know what QH is, other than it is the “best” measure of height. So you choose the weighting factors k1, k2, k3 … k200 that minimize the “error” in your equation determining what QH is over your group of 100 individuals.
Now that you know what the k's are and how to calculate QH, you measure more individuals, but some of them only have 190 bones, and some have 210 bones. You have already determined that k190-k200 are important, so if individuals don't have bones B190-B200, they don't have those components of QH and so are stunted in QH. Similarly since there is no k201-k210, bones B201-B210 are completely unimportant in QH and can be neglected.
As you accumulate data on k's and QH over the decades, you notice that QH is creeping up, that the quintessential height of whole populations is increasing. The k factors that you thought were constant really aren't. But no matter, because you always minimize the error when you calculate your k's and QH, you know that your k's and QH's are reliable in a single cohort, even if they are not reliable between cohorts.
What does “reliable” in this context mean? Does it mean that testing an individual for QH will produce a “true” QH? No, all it means is that in a whole cohort, the average unreliability is distributed across the entire population. That on average, for every high QH there will be a QH that is low, but you can't tell if an individual QH is low, high, or accurate.
Now in the genomic age, you try to fit genetic data to your k's and QH's. But which k's and which QH? The modern ones? Or the less-modern ones? If the different B's have changed in their importance, so too will the importance of the underlying genes (and non-genetic factors) that determine the length the individual B's. But no matter, just pick a model that chooses k factors that minimize the “error” in the QH determination and that “proves” the answer is correct.
I hope you will agree that height is a unitary trait. Does that imply that it is the result of a linear rather than chaotic process? No, to answer that question we have to look further. So again, the issue of whether a trait is unitary is an entirely separate issue than what you are talking about.
Generally speaking, large sample sizes do not eliminate biases. Large sample sizes help reduce random error, not systematic (even unintended) bias.
Nice straw man, I never said the statement was characterized as 100%, I said it was wrongly characterized as consensus. The process by which that statement was developed is a good example of how to elicit groupthink. People said they didn’t sign it because they were afraid it would hurt their careers. We have an example of a researcher who didn’t sign it, and who’s career was hurt and was blocked from publishing because of his intellectual position on these matters. That would be Schonemann. When status depends on conformity to the majority position, that is a perfect setup for groupthink.
I don't care if someone somewhere has described that statement as a consensus (however defined). It's completely irrelevant, and I don't understand why you keep bringing it up.
Some people refused to sign the statement because they knew that being associated with the unpopular (even if widely accepted, tacitly) views of Jensen et al. could hurt their career. To not sign it certainly carried no penalty whatsoever due to the anti-Jensenist bias of the psychological establishment. Lots of psychometricians have had no problem succeeding in their careers despite their vehement anti-Jensen views (whereas in contrast many scholars with un-PC views of racial differences have faced harassment and attempts to have their careers terminated), so if Schonemann was less successful than he had hoped, perhaps it had something to do with his unscholarly tactics and conspiracy-mongering, or perhaps he just was less of a scientist than he thought himself to be. Moreover, despite his protestations to the contrary, Schonemann seems to have got all of his claims published (and refuted) in scholarly journals, so the problem for him was probably just that he couldn't take the intellectual defeat.
You are misreading the paper. It takes large sample sizes (tens of thousands) to avoid bias. There is no evidence that small sample sizes (less than hundreds) are unbiased, other than the unsupported statements of the researchers claiming they are.
I am not misreading anything, and I'm not buying their arguments challenging tons of research just like that. I'll wait for comments and further research by others. They have simply proposed certain hypotheses using certain theorems, made-up data and arbitrary assumptions. If there is a need for a sample of tens of thousands, the bias, if there is any, is tiny. Moreover, the sources of possible bias that they discuss, "acting white" and "stereotype threat", are highly dubious constructs. Finally, as a practical matter, even if there was a tiny bias favoring whites in preemployment testing (a bias in the opposite direction is more likely), it would be meaningless next to the massive advantage affirmative action gives to non-Asian minorities.
Because ‘g’ and IQ are under specified, they can be made to correlate with essentially anything. The way that IQ test questions are selected and incorporated into test scores is to give results that correlate with other IQ tests. If new test questions don’t correlate with old test questions, the new test questions are rejected.
It's pretty much impossible to invent test items that can be thought of as assessing intelligence in some way and that are not simultaneously g-loaded. G-loadings correlate with the complexity and predictive validity of the items, so test designers can't help developing tests that correlate with earlier tests. Otherwise the tests would be useless, with no predictive validity.
g and IQ are not "under specified". They are well-defined, all tests correlate with each other, and g factors derived from different test batteries correlate very highly with each other. Of course they are only estimates in the sense that they do not measure anything in the brain exactly, but then again nothing is exact outside of pure mathematics.
If the tests become obviously bad quickly, what basis is there for thinking the tests are not bad in a shorter time scale?
Because of the high reliability and validity of IQ tests within cohorts. If there are biases within cohorts, they are small and not against any particular ethnic or racial group. In practise, test takers are typically about the same age, so the Flynn effect doesn't matter. Naturally, the validity of the tests would improve if they were renormed more often.
They explicitly say “each IQ gap should not be confused with real (i.e., latent) differences in intelligence.” Which is exactly what you are doing.
Why did you stop quoting? It goes on:
"Only after a proper analysis of measurement invariance of these IQ gaps is conducted can anything be concluded concerning true differencesbetween groups."
And several studies prove that measurement invariance holds in comparisons between blacks and whites. This is what they write:
More importantly, in both B–W studies, it is concluded that the measurement invariance between Blacks and Whites is tenable because the lowest AIC values are found with the factorial invariance models (Dolan, 2000; Dolan & Hamaker, 2001).
It seems that you either don't understand what is said here, or more likely, considering how artfully you failed to quote those passages, you do not want to admit that your position is unsupported. The Flynn effect is irrelevant when it comes to the question of racial differences in IQ.
When the paper you cite says differences in IQ should not be confused with differences in intelligence, what is the HH trying to explain?
Nice straw man, I never said the statement was characterized as 100%, I said it was wrongly characterized as consensus. The process by which that statement was developed is a good example of how to elicit groupthink. People said they didn't sign it because they were afraid it would hurt their careers. We have an example of a researcher who didn't sign it, and who's career was hurt and was blocked from publishing because of his intellectual position on these matters. That would be Schonemann. When status depends on conformity to the majority position, that is a perfect setup for groupthink.
The only genetic connections we are confident are related to mental retardation are single gene problems that actually do “fall right out” of a genome wide survey. The gene researchers expected genes associated with disorders characterized as highly hereditary to “fall right out” of genome wide surveys before those surveys were done. They didn't. The reason they don't is because genetic systems are very complicated. They comprise multiple non-linear coupled parameters. They are not “additive”, or “linear”, or reducible to anything like a X% heritability. Non-linear coupled systems are inherently chaotic and exhibit what is called the butterfly effect. They can be extremely sensitive to initial conditions, such as the conditions in utero during the first trimester where the same genome can give rise to a normal and an anencephalic infant.
The whole idea of 'g' and IQ is predicated on what is called “intelligence” being a linear function of stuff. Trying to model chaotic systems with linear models will always fail. What “percentage” genetic is the anencephalopathy that one MZT exhibits when the other MZT does not?
You are misreading the paper. It takes large sample sizes (tens of thousands) to avoid bias. There is no evidence that small sample sizes (less than hundreds) are unbiased, other than the unsupported statements of the researchers claiming they are.
Because 'g' and IQ are under specified, they can be made to correlate with essentially anything. The way that IQ test questions are selected and incorporated into test scores is to give results that correlate with other IQ tests. If new test questions don't correlate with old test questions, the new test questions are rejected.
This isn't a question of “belief”, it is a question of data and do we make policy based on belief or based on data.
You haven't addressed the Flynn Effect. The paper you cited and quoted from (Wicherts et al) says:
“Each comparison of groups should be investigated separately. IQ gaps between cohorts do not teach us anything about IQ gaps between contemporary groups, except that each IQ gap should not be confused with real (i.e., latent) differences in intelligence. Only after a proper analysis of measurement invariance of these IQ gaps is conducted can anything be concluded concerning true differences between groups.”
“Whereas implications of the Flynn effect for B–W differences appear small, the implications for intelligence testing, in general, are large. That is, the Flynn effect implies that test norms become obsolete quite quickly (Flynn, 1987).”
where you stop your quotation. If you continue the quotation it says:
“More importantly, however, the rejection of factorial invariance within a time period of only a decade implies that even subtest score interpretations become obsolete. Differential gains resulting in measurement bias, for example, imply that an overall test score (i.e., IQ) changes in composition. The effects on the validity of intelligence tests are unknown, but one can easily imagine that the factors that cause bias over the years also influence within-cohort differences. Further research on the causes of the artifactual gains is clearly needed.”
“The overall conclusion of the present paper is that factorial invariance with respect to cohorts is not tenable. Clearly, this finding requires replication in other data sets. However, if this finding proves to be consistent, it should have implications for explanations of the Flynn effect. The fact that the gains cannot be explained solely by increases at the level of the latent variables (common factors), which IQ tests purport to measure, should not sit well with explanations that appeal solely to changes at the level of the latent variables.”
What they are saying is that the Flynn Effect is at least partly artifactual due to the tests being bad and not measuring what they claim to be measuring (latent variables) from which IQ can be inferred. If the tests become obviously bad quickly, what basis is there for thinking the tests are not bad in a shorter time scale?
They explicitly say “each IQ gap should not be confused with real (i.e., latent) differences in intelligence.” Which is exactly what you are doing.
When the paper you cite says differences in IQ should not be confused with differences in intelligence, what is the HH trying to explain? Differences in how people do on bad tests that have been designed to correlate with other bad tests?
"If the heterosis idea was actually correct, then because Africa has the vast majority of human genetic diversity, the intelligence of the population of Africa should be much higher than everywhere else in the world, and should have been since human speciation. "No, you are making too big an inference. denis was suggesting that heterosis is responsible for the increase over recent generations. That does NOT imply that heterosis explains all variation in intelligence. So his theory would imply that Africans have more intelligence THAN THEY WOULD HAVE HAD if they had not been so diverse.
I'm no expert on the thought of hereditarians, but from what I've read they tend to promote things like breastfeeding & micronutrients. As far as I know there are no hereditarian researchers who believe in a 100% GENETIC explanation. There are nurturists (generally not psychometricians) who try to clamp down on any genetic explanations, but nobody does the same for environmental influence (though some think there are no "shared environment" effects among normal middle class first-world families, still leaving room for peer effects).
Uterine environments are an interesting subject and I think there should be more research into them. Greg Cochran (categorizable as a hereditarian, though he is a physicist-turned-anthropologist) thinks schizophrenia is caused by an infection, like narcolepsy. There have been a lot of nonsensical claims about the causes of autism, so I'd need some cites for your listed causes before I go repeating them credulously (as is normally my wont). The latest hypothesis I've heard is that ultrasounds explain its prevalence among wealthier parents. Who knows.
In my earlier comment I had meant to reference personality. That's another branch of psychometrics, and nobody thinks there is a unitary "g" factor for it (I think the five-factor theory is most popular now). Nevertheless, personality seems significantly heritable. So there's another example of unitary vs heritable being separable.
The major rivals to the theory of general intelligence are Sternberg's "triarchic" theory and Gardner's "multiple intelligences". They believe, as with wet earwax and tasting PTC, there is something real there (though as with all analogies we shouldn't try to stretch it too much). They simply believe that intelligence is a more multifaceted thing. Hereditarianism could still apply just the same under their theories (though ceteris paribus one would expect lower correlations). You seem to be suggesting that intelligence tests don't really measure anything at all, which is a position I haven't heard anyone with a psychometric background make since the tests have a considerable amount of predictive power. If you haven't yet grasped the distinction between whether a trait is unitary vs heritable, I advise you to do more reading up on the subject from sources other than the ones that have plainly failed to impart that distinction.
I have seen the statement published in the WSJ wrongly characterized as a “consensus”. Mainstream and consensus are (to me), very different things.
Neither the statement nor anyone in this discussion has claimed that 100% of intelligence researchers agree with the statement, so that's completely irrelevant.
First, there is no unitary “g” or other unitary measure of intelligence that can be measured as with intelligence testing. The tests to determine such a thing are all fatally flawed. The concept of “g” is underspecified.
Firstly, even if factor analysis had never been invented, there would be IQ tests and the heritability of IQ could be studied.
Secondly, g extracted from any sufficiently diverse battery of tests is pretty much the same, so that, for example, the rank ordering of people in "g intelligence" is similar regardless of the test used.
Second, the Flynn Effect shows large changes in IQ in whole populations not due to genetics. If there are large differences in IQ between populations that are not due to genetics, what basis is there for assuming that smaller differences are due to genetics?
The same basis that there is for assuming that human height differs between individuals, populations and sexes partly for genetic reasons despite the increases in average height during recent generations. (Now, I understand that you don't believe that even height is heritable, but that's just too ridiculous an idea to merit further discussion.) As I have demonstrated, Flynn gains and the black-white IQ difference are qualitatively different, and the Flynn effect has no causal connection to the b-w IQ gap.
Third. Now that we have actual genetic data on different populations, there has been no discovery of genes that explain differences in intelligence, not between individuals, not between populations. If intelligence were primarily genetic, this would fall right out of the data.
That shows that you're just talking out of your ass, without having done any reading whatsoever on the topic. If there is one thing we know about genome-wide association studies and other genetic methods, it's that nothing "falls right out of the data". Results like this, this, and this are the fruit of years of work. Moreover, I am not aware of any studies of race and IQ where these methods are used. However, it's incorrect to say that we do not know of any genetic differences that influence intelligence. In fact, we know plenty that cause mental retardation.
Fourth, The very recent paper I linked to, shows that the sample sizes that have been used are inadequate to be unbiased.
Your enormous confidence in the tall claims challenging tons of replicated results made in one recent paper suggests that you're more interested in defending your poorly-thought out claims that finding out the truth. That study uses mocked-up data, and the fact that huge sample sizes are needed for there to be any bias even in such an artificial scenario suggests that even if the tests are biased, the bias is tiny (and likely disfavors whites).
If there is no such thing as ‘g’, what is it that the HH is trying to explain? If you can’t define or measure “intelligence”, how can you measure variance in intelligence. If you can’t reliably measure intelligence, what exactly are you trying to explain via genetics?
IQ tests are highly reliable in the sense of test-retest reliability, and the results of different IQ tests are highly intercorrelated. They are a valid measure of intelligence because of their predictive validity, even if different tests measuring other aspects of cognition could putatively be developed. IQ is certainly more reliable than, say, the diagnoses of many diseases, and there are plenty of GWA studies of diseases. I believe we will learn about the genetic basis of intelligence sooner rather than later.
"So why do you accept the conclusion if the argument is incomplete?"
Your argument is absolutist in tone though, by your own logic we shouldn't accept evolution per se due to the fact the chain of transitional fossils has not been 'completed' yet.
Nice to you see smugly attacking Catholics and Homeopathy practitioners on your blog though. Isn't it funny how egalitarians always pick the easiest targets to beat over the head? Not a single one would ever challenge the real dominant paradigms that rule our age, the real taboos of racial egalitarianism and how any inequality of outcome is invariably down to some kind of 'phantom discrimination' and can be corrected with social engineering.
Far from it, you support these paradigms and gleefully engage in the kind of character assassinations that followed Watson's comments years ago, or Lahn's research into Microcephalin and so on.
Fuck Zionism by the way. If you want to talk about 'religious evil' just read the damned Talmud, I've never seen so much hateful supremacist filth in all my life, the Babylonian Talmud is like Mein Kampf on steroids.
How about this link
http://www.ncbi.nlm.nih.gov...
The concordance rate for full siblings is in the 3-6% range.
http://resources.metapress....
There is cross-talk between twin fetuses in utero. In cattle, there is what is known as a “free martin”, the female twin of a male calf. A free martin is usually always sterile.
The increased autism in male-male DZT is probably due to higher levels of testosterone.
Also, I only have access to the abstract for the ncbi paper. It only references the concordances for MZ and DZ twins (by gender). Could you give the concordance for full-siblings here?
That's pretty neat, Jason. I would have assumed that the variance in Mendelian segregation, as well as the portion of the genome affecting height, was small enough that it would be hard to find anything that way.
Like the commenter "Miko" there, I am outside my domain competence. But in browsing the paper they only reference gender & age as assumed to have linear effects. Also, daedaulus2u, your plos link is broken.
Speaking of pre-natal effects, I have occasionally heard (such as from the second video here) that crack babies don't actually turn out worse than comparable children not exposed to crack. That's surprising to me, do any of you know much about it?
Uh Jason, you should look at the comment in the PloS article you linked to. It is not “assumption-free”, it is “fewer assumptions”. As was ably pointed out, they still assume that gene effects are not interactive other than being linearly additive. We know that genes have effects that are not linearly additive. That analysis assumes that those interactive effects are all negligible. We know they are not.
If you look at a gene like MeCP2, we now know that loss of it causes Rett Syndrome in females and is fatal in males (pretty strong rescue of the phenotype by interaction with an X chromosome which happens to be inactive in the cells that are rescued). The MeCP2 protein binds to methylated DNA, and so MeCP2 regulates how DNA that has been epigenetically programmed by DNA methylation is read out. MeCP2 doesn't have any direct effects on phenotype, the only effects are secondary, mediated through differential effects on the transcription of other genes.
DNA methyl transferases also only have effects through differential regulation of other genes. SNPs of DNA methyl transferases do seem to have effects on intelligence.
http://www.plosone.org/arti...
Presumably those DNA methyl transferases are having effects through their normal function, that of methylating DNA and so affecting how it is transcribed.
How about genes for transcription factors? They don't have direct effects either, only effects mediated through differential transcription of other genes. Something like 5% of the human genome is transcription factors. They only have effects mediated through gene-gene interactions.
People have actually looked for the genes that supposedly cause intelligence to be inherited. Have they been found yet? No, they have not. Until the specific genes are found and reliably identified as influencing intelligence and under what environmental conditions, their existence remains hypothetical. Due to the level of effort put into finding such genes, and the essentially complete lack of success in finding them, to me their existence is becoming increasingly speculative.
Pretending that models of inheritance that ignore gene-gene interactions have risen to the level of “fact” is disingenuous, or as you would say, “a lie”.
Uh, we know a lot more than was known in 1950. We now know that the actual genetic material is DNA, and we now know a lot about how it is replicated, stored, regulated, epigenetically programmed, silenced and transcribed. We have actual data on similarities of twins vs siblings, and fraternal twins. We now know that fraternal twins have a higher concordance for autism than do full siblings.
http://www.ncbi.nlm.nih.gov...
Sharing an in utero environment does make fraternal twins more similar along the autism spectrum than full siblings. If fraternal twins have a higher incidence of autism, then presumably there is something about their shared in utero environment that affects the normal process of neurodevelopment and causes autism. Presumably there are also sub-clinical effects of what ever that in utero environmental process is that produce effects less than autism. Presumably those effects might include something like intelligence.
TGGP: "we can distinguish between identical twins, fraternal twins, and unrelated children raised together (or apart). This is how measures of heritability are obtained"
daedalus2u: "The major “data” linking genes to intelligence is from a naïve examination of monozygous twins. Monozygous twins share a lot more than just a genome, they also shared pretty much the same in utero environment where their brains grew from a single cell to 100,000,000,000 cells, a factor of 10^11. "
Evidence for the high heritability of intelligence, and other traits, is extensive, multifaceted, and unimpeachable. It is not based on one method, but on numerous, mutually reinforcing methods. Whatever working assumptions behavior genetic methods may have started with, have long since entered the realm of supported fact.
Genetic denialism is quackier than ever. We don't even need different levels of kinship anymore (e.g. identical twins, cousins, etc), we can now demonstrate substantial heritability using nothing but full siblings! Welcome to the Post-Genomic Era; take this flyer and please remember to get your hand stamped at the door.
And not that it matters (since the assertion it was tied to was a lie), but sharing a prenatal environment overwhelmingly makes siblings less alike, not more alike.
Yes, when I said "imperfectly measured" that was a reference to the flawed nature of the procedure. I think you now exaggerates its problems though. As Robyn Dawes explained in "Rational Choice in an Uncertain World", statistical models with randomly chosen coefficients do surprisingly well. The important thing is that they have the right sign.
"The HH requires modeling intelligence as a linear combination of genes and environmental influences"The hereditarian hypothesis is that a substantial portion of the variation in intelligence is due to genetic variation (or alternatively, that the same is true for variation among population groups). If IQ was attributable purely to a massively complex interaction of countless genes, hereditarianism would be obviously correct. If IQ was purely the result of a simple linear combination of environmental factors, hereditarianism would be obviously incorrect.
"A genotype growing up in one environment may have a high IQ phenotype and in another environment may have a low IQ phenotype. One environment may produce a high IQ in one genotype and a low IQ in another."That can be the case even in a linear interaction model. You could correct it by specifying that in the first sentence you are speaking relative to a different genotype but the same environment, and vice versa for the second sentence. Unfortunately, we don't yet know enough about genetics to make such comparisons. We can observe that there are still large amounts of variation for identical twins with very similar environments, but that's compatible with a linear interaction model in which random noise plays a large role.
On stress and measurement: we can distinguish between identical twins, fraternal twins, and unrelated children raised together (or apart). This is how measures of heritability are obtained (though they can only directly shed light on contributions to within-group variation rather than between group variation, biracial children perhaps shedding light on the latter). We don't have equivalent comparable classes for stress.
But if there are ten-dollar bills on the sidewalk researchers have overlooked in the form of unmeasured outcomes, perhaps you should start contacting researchers and offering suggestions. I often email professors out of the blue and they can be surprisingly receptive. There is a subfield of psychology devoted to childhood development, and I would expect that is your best bet.
You do appreciate that the procedure I outlined for calculating height from bone length is flawed, even though height is (essentially) a linear combination of the lengths of certain bones? The coefficients are pretty close to either zero or one. Zero for bones like ribs and 1 for bones like the femur. If a calculating procedure like that fails for something as simple as height and bone length, why would anyone expect such a procedure to work for something as complex and as poorly defined as intelligence? It would virtually certainly work less well.
Intelligence is not a linear combination of anything. It is highly non-linear. The problem is that the interactions between the genes and the environment are non-linear and the interactions are coupled.
The HH requires modeling intelligence as a linear combination of genes and environmental influences of the form (where G1 is gene 1 and EI2 is environmental influence 2 and g1 is the constant that determines the effect of G1):
IQ = g1(G1) + g2(G2) + … + ei1(EI1) + ei2(EI2) + …
In reality, the g's and e's interact with each other and with themselves. The first coefficient isn't a constant, it is a non-linear function of all the G's and all the E's and all their mutual interactions that collectively influence the effect of G1 on intelligence.
It is a non-linear function with at least hundreds of genetic terms and likely thousands of environmental terms. These functions are different for each different tissue compartment and cell type that has an influence on neurodevelopment and brain function, heart; liver, kidneys, immune system and vary over time and exhibit hysteresis. It is also epigenetically programmed in utero and maybe in parental gametes.
I appreciate that the actual complexity of the genetic influence on intelligence is too complex to model given our current understanding of genetics and environmental influences. The HH requires neglecting all this complexity and ignoring all gene-gene interactions, all environment-environment interactions and all gene-environment interactions. We know that is not correct.
Let me restate it a different way; it is extremely likely that genes that affects IQ can have both positive and negative effects on IQ depending on other genes and on the environment. The effects of a particular gene on IQ are not just “small”, they can be either positive or negative depending on other things. A genotype growing up in one environment may have a high IQ phenotype and in another environment may have a low IQ phenotype. One environment may produce a high IQ in one genotype and a low IQ in another. Environments that are nominally “the same” can produce completely different results even in the same genome, as in the case of MZT being discordant for anencephalopathy.
Many of the genes that affect IQ are active in utero, when the brain grows from a single cell to 10^11 cells. Epigenetic programming of DNA in utero is known to affect animal behaviors as adults. Exposure to stress in utero is known to affect adult behaviors in animals and in humans. Are there differential levels of stress between the B/W population? Are pregnant white women exposed to more or less stress than pregnant black women? What is the basis for ignoring those differential levels of stress? Because stress levels are hard to measure? Gene frequency hasn't been measured at all, but the HH has no difficulty attributing large effects to unmeasured and unknown genes while completely neglecting in utero stress which is known to be different, and could be measured. Certainly outcomes attributable to stressful pregnancy are known (low birth weight) and could be measured but are not.
Yes, height can be thought of as the sum of many smaller components, but it is still a unitary trait: it can be accurately expressed with a single scalar. In your analogy there is a trait which is real and unitary in nature but imperfectly measured.
I have never heard anyone say that IQ is the result of a single "IQ gene". The standard take among hereditarians I've heard is that there are a very large number of genes with individually small effects. The addition of such large numbers of small factors results in normal distributions. There are also of course some genes with large effects that give rise to retardation.
Height is not a unitary measure of any physiological primitive. Height is the sum of the lengths of the femur, the tibia, the individual bones in the spine, the diameter of the skull, the thickness of the scalp, the hip, the foot bones, the thickness of the foot pad, plus the thickness of cartilage between each pair of bones.
The way that the length of a bone like the femur is determined, is primarily when the growth plates stop growing, stop adding length to the ends of the bone. That growth is determined locally, by the cells that comprise the growth plate. Skull diameter is mostly determined by brain size, but skull shape matters too. Should we expect there to be a single unitary “height gene” that synchronously controls the proportionate size of the height component of each bone independent of other physiological parameters? No, we shouldn't.
Suppose you didn't have the ability to measure distances longer than 1 meter, but still wanted to find the “true” or “quintessential height” (QH) of an individual by taking a weighted sum of the lengths of the individual bones. Since you are not quite sure how the different bone lengths will add up, or which bones will be important, you decide to take a weighted sum of all the bones you can measure and determine which ones are most important.
Assuming you can measure 200 bones, you measure those individual bone lengths of 100 individuals and then determine the weighting factors to use in weighting the individual bone lengths to quintessential height. You don't know what quintessential height actually is, so you use the weighting factors that give you the “best” fit but don't know exactly what weight to put to each of them, so you solve for the weighting factors that will give you the “best” overall fit. You now have equations of the form
k1*B1 + k2*B2 + k3*B3 + … + k200*B200 = QH (quintessential height)
You still don't know what QH is, other than it is the “best” measure of height. So you choose the weighting factors k1, k2, k3 … k200 that minimize the “error” in your equation determining what QH is over your group of 100 individuals.
Now that you know what the k's are and how to calculate QH, you measure more individuals, but some of them only have 190 bones, and some have 210 bones. You have already determined that k190-k200 are important, so if individuals don't have bones B190-B200, they don't have those components of QH and so are stunted in QH. Similarly since there is no k201-k210, bones B201-B210 are completely unimportant in QH and can be neglected.
As you accumulate data on k's and QH over the decades, you notice that QH is creeping up, that the quintessential height of whole populations is increasing. The k factors that you thought were constant really aren't. But no matter, because you always minimize the error when you calculate your k's and QH, you know that your k's and QH's are reliable in a single cohort, even if they are not reliable between cohorts.
What does “reliable” in this context mean? Does it mean that testing an individual for QH will produce a “true” QH? No, all it means is that in a whole cohort, the average unreliability is distributed across the entire population. That on average, for every high QH there will be a QH that is low, but you can't tell if an individual QH is low, high, or accurate.
Now in the genomic age, you try to fit genetic data to your k's and QH's. But which k's and which QH? The modern ones? Or the less-modern ones? If the different B's have changed in their importance, so too will the importance of the underlying genes (and non-genetic factors) that determine the length the individual B's. But no matter, just pick a model that chooses k factors that minimize the “error” in the QH determination and that “proves” the answer is correct.
I hope you will agree that height is a unitary trait. Does that imply that it is the result of a linear rather than chaotic process? No, to answer that question we have to look further. So again, the issue of whether a trait is unitary is an entirely separate issue than what you are talking about.
Generally speaking, large sample sizes do not eliminate biases. Large sample sizes help reduce random error, not systematic (even unintended) bias.
Nice straw man, I never said the statement was characterized as 100%, I said it was wrongly characterized as consensus. The process by which that statement was developed is a good example of how to elicit groupthink. People said they didn’t sign it because they were afraid it would hurt their careers. We have an example of a researcher who didn’t sign it, and who’s career was hurt and was blocked from publishing because of his intellectual position on these matters. That would be Schonemann. When status depends on conformity to the majority position, that is a perfect setup for groupthink.
I don't care if someone somewhere has described that statement as a consensus (however defined). It's completely irrelevant, and I don't understand why you keep bringing it up.
Some people refused to sign the statement because they knew that being associated with the unpopular (even if widely accepted, tacitly) views of Jensen et al. could hurt their career. To not sign it certainly carried no penalty whatsoever due to the anti-Jensenist bias of the psychological establishment. Lots of psychometricians have had no problem succeeding in their careers despite their vehement anti-Jensen views (whereas in contrast many scholars with un-PC views of racial differences have faced harassment and attempts to have their careers terminated), so if Schonemann was less successful than he had hoped, perhaps it had something to do with his unscholarly tactics and conspiracy-mongering, or perhaps he just was less of a scientist than he thought himself to be. Moreover, despite his protestations to the contrary, Schonemann seems to have got all of his claims published (and refuted) in scholarly journals, so the problem for him was probably just that he couldn't take the intellectual defeat.
You are misreading the paper. It takes large sample sizes (tens of thousands) to avoid bias. There is no evidence that small sample sizes (less than hundreds) are unbiased, other than the unsupported statements of the researchers claiming they are.
I am not misreading anything, and I'm not buying their arguments challenging tons of research just like that. I'll wait for comments and further research by others. They have simply proposed certain hypotheses using certain theorems, made-up data and arbitrary assumptions. If there is a need for a sample of tens of thousands, the bias, if there is any, is tiny. Moreover, the sources of possible bias that they discuss, "acting white" and "stereotype threat", are highly dubious constructs. Finally, as a practical matter, even if there was a tiny bias favoring whites in preemployment testing (a bias in the opposite direction is more likely), it would be meaningless next to the massive advantage affirmative action gives to non-Asian minorities.
Because ‘g’ and IQ are under specified, they can be made to correlate with essentially anything. The way that IQ test questions are selected and incorporated into test scores is to give results that correlate with other IQ tests. If new test questions don’t correlate with old test questions, the new test questions are rejected.
It's pretty much impossible to invent test items that can be thought of as assessing intelligence in some way and that are not simultaneously g-loaded. G-loadings correlate with the complexity and predictive validity of the items, so test designers can't help developing tests that correlate with earlier tests. Otherwise the tests would be useless, with no predictive validity.
g and IQ are not "under specified". They are well-defined, all tests correlate with each other, and g factors derived from different test batteries correlate very highly with each other. Of course they are only estimates in the sense that they do not measure anything in the brain exactly, but then again nothing is exact outside of pure mathematics.
If the tests become obviously bad quickly, what basis is there for thinking the tests are not bad in a shorter time scale?
Because of the high reliability and validity of IQ tests within cohorts. If there are biases within cohorts, they are small and not against any particular ethnic or racial group. In practise, test takers are typically about the same age, so the Flynn effect doesn't matter. Naturally, the validity of the tests would improve if they were renormed more often.
They explicitly say “each IQ gap should not be confused with real (i.e., latent) differences in intelligence.” Which is exactly what you are doing.
Why did you stop quoting? It goes on:
"Only after a proper analysis of measurement invariance of these IQ gaps is conducted can anything be concluded concerning true differencesbetween groups."
And several studies prove that measurement invariance holds in comparisons between blacks and whites. This is what they write:
More importantly, in both B–W studies, it is concluded that the measurement invariance between Blacks and Whites is tenable because the lowest AIC values are found with the factorial invariance models (Dolan, 2000; Dolan & Hamaker, 2001).
It seems that you either don't understand what is said here, or more likely, considering how artfully you failed to quote those passages, you do not want to admit that your position is unsupported. The Flynn effect is irrelevant when it comes to the question of racial differences in IQ.
When the paper you cite says differences in IQ should not be confused with differences in intelligence, what is the HH trying to explain?
That's not what they say. Learn to read.
Nice straw man, I never said the statement was characterized as 100%, I said it was wrongly characterized as consensus. The process by which that statement was developed is a good example of how to elicit groupthink. People said they didn't sign it because they were afraid it would hurt their careers. We have an example of a researcher who didn't sign it, and who's career was hurt and was blocked from publishing because of his intellectual position on these matters. That would be Schonemann. When status depends on conformity to the majority position, that is a perfect setup for groupthink.
The only genetic connections we are confident are related to mental retardation are single gene problems that actually do “fall right out” of a genome wide survey. The gene researchers expected genes associated with disorders characterized as highly hereditary to “fall right out” of genome wide surveys before those surveys were done. They didn't. The reason they don't is because genetic systems are very complicated. They comprise multiple non-linear coupled parameters. They are not “additive”, or “linear”, or reducible to anything like a X% heritability. Non-linear coupled systems are inherently chaotic and exhibit what is called the butterfly effect. They can be extremely sensitive to initial conditions, such as the conditions in utero during the first trimester where the same genome can give rise to a normal and an anencephalic infant.
The whole idea of 'g' and IQ is predicated on what is called “intelligence” being a linear function of stuff. Trying to model chaotic systems with linear models will always fail. What “percentage” genetic is the anencephalopathy that one MZT exhibits when the other MZT does not?
You are misreading the paper. It takes large sample sizes (tens of thousands) to avoid bias. There is no evidence that small sample sizes (less than hundreds) are unbiased, other than the unsupported statements of the researchers claiming they are.
Because 'g' and IQ are under specified, they can be made to correlate with essentially anything. The way that IQ test questions are selected and incorporated into test scores is to give results that correlate with other IQ tests. If new test questions don't correlate with old test questions, the new test questions are rejected.
This isn't a question of “belief”, it is a question of data and do we make policy based on belief or based on data.
You haven't addressed the Flynn Effect. The paper you cited and quoted from (Wicherts et al) says:
“Each comparison of groups should be investigated separately. IQ gaps between cohorts do not teach us anything about IQ gaps between contemporary groups, except that each IQ gap should not be confused with real (i.e., latent) differences in intelligence. Only after a proper analysis of measurement invariance of these IQ gaps is conducted can anything be concluded concerning true differences between groups.”
“Whereas implications of the Flynn effect for B–W differences appear small, the implications for intelligence testing, in general, are large. That is, the Flynn effect implies that test norms become obsolete quite quickly (Flynn, 1987).”
where you stop your quotation. If you continue the quotation it says:
“More importantly, however, the rejection of factorial invariance within a time period of only a decade implies that even subtest score interpretations become obsolete. Differential gains resulting in measurement bias, for example, imply that an overall test score (i.e., IQ) changes in composition. The effects on the validity of intelligence tests are unknown, but one can easily imagine that the factors that cause bias over the years also influence within-cohort differences. Further research on the causes of the artifactual gains is clearly needed.”
“The overall conclusion of the present paper is that factorial invariance with respect to cohorts is not tenable. Clearly, this finding requires replication in other data sets. However, if this finding proves to be consistent, it should have implications for explanations of the Flynn effect. The fact that the gains cannot be explained solely by increases at the level of the latent variables (common factors), which IQ tests purport to measure, should not sit well with explanations that appeal solely to changes at the level of the latent variables.”
What they are saying is that the Flynn Effect is at least partly artifactual due to the tests being bad and not measuring what they claim to be measuring (latent variables) from which IQ can be inferred. If the tests become obviously bad quickly, what basis is there for thinking the tests are not bad in a shorter time scale?
They explicitly say “each IQ gap should not be confused with real (i.e., latent) differences in intelligence.” Which is exactly what you are doing.
When the paper you cite says differences in IQ should not be confused with differences in intelligence, what is the HH trying to explain? Differences in how people do on bad tests that have been designed to correlate with other bad tests?
"If the heterosis idea was actually correct, then because Africa has the vast majority of human genetic diversity, the intelligence of the population of Africa should be much higher than everywhere else in the world, and should have been since human speciation. "No, you are making too big an inference. denis was suggesting that heterosis is responsible for the increase over recent generations. That does NOT imply that heterosis explains all variation in intelligence. So his theory would imply that Africans have more intelligence THAN THEY WOULD HAVE HAD if they had not been so diverse.
I'm no expert on the thought of hereditarians, but from what I've read they tend to promote things like breastfeeding & micronutrients. As far as I know there are no hereditarian researchers who believe in a 100% GENETIC explanation. There are nurturists (generally not psychometricians) who try to clamp down on any genetic explanations, but nobody does the same for environmental influence (though some think there are no "shared environment" effects among normal middle class first-world families, still leaving room for peer effects).
Uterine environments are an interesting subject and I think there should be more research into them. Greg Cochran (categorizable as a hereditarian, though he is a physicist-turned-anthropologist) thinks schizophrenia is caused by an infection, like narcolepsy. There have been a lot of nonsensical claims about the causes of autism, so I'd need some cites for your listed causes before I go repeating them credulously (as is normally my wont). The latest hypothesis I've heard is that ultrasounds explain its prevalence among wealthier parents. Who knows.
In my earlier comment I had meant to reference personality. That's another branch of psychometrics, and nobody thinks there is a unitary "g" factor for it (I think the five-factor theory is most popular now). Nevertheless, personality seems significantly heritable. So there's another example of unitary vs heritable being separable.
The major rivals to the theory of general intelligence are Sternberg's "triarchic" theory and Gardner's "multiple intelligences". They believe, as with wet earwax and tasting PTC, there is something real there (though as with all analogies we shouldn't try to stretch it too much). They simply believe that intelligence is a more multifaceted thing. Hereditarianism could still apply just the same under their theories (though ceteris paribus one would expect lower correlations). You seem to be suggesting that intelligence tests don't really measure anything at all, which is a position I haven't heard anyone with a psychometric background make since the tests have a considerable amount of predictive power. If you haven't yet grasped the distinction between whether a trait is unitary vs heritable, I advise you to do more reading up on the subject from sources other than the ones that have plainly failed to impart that distinction.
I have seen the statement published in the WSJ wrongly characterized as a “consensus”. Mainstream and consensus are (to me), very different things.
Neither the statement nor anyone in this discussion has claimed that 100% of intelligence researchers agree with the statement, so that's completely irrelevant.
First, there is no unitary “g” or other unitary measure of intelligence that can be measured as with intelligence testing. The tests to determine such a thing are all fatally flawed. The concept of “g” is underspecified.
Firstly, even if factor analysis had never been invented, there would be IQ tests and the heritability of IQ could be studied.
Secondly, g extracted from any sufficiently diverse battery of tests is pretty much the same, so that, for example, the rank ordering of people in "g intelligence" is similar regardless of the test used.
Second, the Flynn Effect shows large changes in IQ in whole populations not due to genetics. If there are large differences in IQ between populations that are not due to genetics, what basis is there for assuming that smaller differences are due to genetics?
The same basis that there is for assuming that human height differs between individuals, populations and sexes partly for genetic reasons despite the increases in average height during recent generations. (Now, I understand that you don't believe that even height is heritable, but that's just too ridiculous an idea to merit further discussion.) As I have demonstrated, Flynn gains and the black-white IQ difference are qualitatively different, and the Flynn effect has no causal connection to the b-w IQ gap.
Third. Now that we have actual genetic data on different populations, there has been no discovery of genes that explain differences in intelligence, not between individuals, not between populations. If intelligence were primarily genetic, this would fall right out of the data.
That shows that you're just talking out of your ass, without having done any reading whatsoever on the topic. If there is one thing we know about genome-wide association studies and other genetic methods, it's that nothing "falls right out of the data". Results like this, this, and this are the fruit of years of work. Moreover, I am not aware of any studies of race and IQ where these methods are used. However, it's incorrect to say that we do not know of any genetic differences that influence intelligence. In fact, we know plenty that cause mental retardation.
Fourth, The very recent paper I linked to, shows that the sample sizes that have been used are inadequate to be unbiased.
Your enormous confidence in the tall claims challenging tons of replicated results made in one recent paper suggests that you're more interested in defending your poorly-thought out claims that finding out the truth. That study uses mocked-up data, and the fact that huge sample sizes are needed for there to be any bias even in such an artificial scenario suggests that even if the tests are biased, the bias is tiny (and likely disfavors whites).
If there is no such thing as ‘g’, what is it that the HH is trying to explain? If you can’t define or measure “intelligence”, how can you measure variance in intelligence. If you can’t reliably measure intelligence, what exactly are you trying to explain via genetics?
IQ tests are highly reliable in the sense of test-retest reliability, and the results of different IQ tests are highly intercorrelated. They are a valid measure of intelligence because of their predictive validity, even if different tests measuring other aspects of cognition could putatively be developed. IQ is certainly more reliable than, say, the diagnoses of many diseases, and there are plenty of GWA studies of diseases. I believe we will learn about the genetic basis of intelligence sooner rather than later.