Suppose you wanted to believe, as I do, that intelligence and vague ‘racial groups’ are, on the whole, unrelated from a long-run perspective. What would you then have to believe about genetics and IQ, as well as the long-run effects of socio-economic circumstances on IQ to rationalise the overwhelming evidence that there is a strong correlation between IQ and ‘race’? That evidence is extensively discussed in this excellent blog by Ken Parish, alluded to in several references on this wikipedia page into IQ heritability, and pushed hard in this book I co-reviewed recently? I will argue in the below that the main thing you must then dismiss is the hypothesis that a strong correlation between the outcomes befalling identical twins ‘demonstrates’ the importance of shared genes. In turn, that forces you to re-evaluate a lot of medical and biological studies in many other fields.
Consider the basic evidence that the ‘Bell Curve’ adherents have to argue their case: they argue that starting from the darkest places in Africa with IQs around 75 you move away from the equator and see an increase in both wealth and IQ, passing over the US with an IQ of 100, ending up with Japan and South Korea where average IQs are around 110 and health and wealth are concomitantly high. Within rich countries too, the basic empirical relation is that the whiter the ethnic group, the higher the IQ scores. And everything that is desirable is positively related with IQ, including length of life, wealth, low crime, etc. The ‘Bell Curve’ adherents point to all these correlations and say they are causal and that, furthermore, there is a genetic component to IQ since IQ is strongly heritable within families: the IQ of identical twins who were separated at birth turns out to be pretty close, usually taken as evidence of genetic causality.
So the mountain to climb is to rationalise all those correlations without buying into genetic causality at the level of ethnic groups, loosely labelled as ‘race’. Find below my best efforts, allowing for the fact that I am of course mainly a consumer of this literature, not a fully up-skilled producer!
The simplest put-down is to say that IQ is not intelligence but measures a culturally-specific skill that, quite naturally, would be higher in countries oriented towards teaching those skills and valuing them. This put-down can be further motivated by saying that IQ is really just defined by its measurement, and is thus defined as ‘the skills that predict school scores’, which does not by necessity capture anything that is fixed for a person or that is independent of culture. Indeed, it is basically impossible to measure intelligence free of culture since from our very earliest days we direct our learning towards that which is valued by our surroundings. Education is similarly not intended to be free of culture and so the thing that predicts it will also not be free of it. So the first put-down is that intelligence cannot be measured free of culture and hence that comparing across ethnicities or racial groups is impossible to start with.
This put-down, which you see a lot, is not really sustainable. For one, the notion of ‘intelligence free of culture’ is a form of ‘humans without humanity’: a contradiction in terms. Hence intelligence is also specific to cultures, meaning that there is no such thing as ‘equal intelligence outside cultures’ either, so whilst it is true that IQ measures ‘ability to do tests valued in the rich world’ , that does not mean that one cannot talk about the heritability of being able to do those ‘Western schooling calibrated tests’.
Moreover, it turns out that high IQ in one culture does not translate into low IQ in another, a point also pushed by the very politically-correct James Flynn in his recent book: people with higher IQs on average are better and can more quickly learn almost every mental task you want to throw at them. So IQ might be specific to a particular test but someone who is 30 points ahead in IQ probably beats someone else at their own mental game as well. Worse, ethnic groups and species with higher IQs do on average seem to have bigger brains, as you would expect. It really is hard to believe that bigger brains do not mean one is smarter on average (accounting for body size). Worst of all, even in societies where one can argue that there are no major current cultural differences between groups, such as the various groups that have lived in the US for over 100 years and that hence now all speak English and have been to school for 3 generations, still show these strong average IQ differences between the ethnic groups.
A similar unconvincing put-down is to say that IQ cannot possibly have causal effects on wealth, crime, etc. because other things than intelligence cause those. Simply by the fact that IQ is the ‘thing that predicts education’ one would thus have to argue that education has no causal effect on wealth, health, crime, etc., because higher IQ on average does cause better education (the higher your IQ the further you can rise through the stages of education). The put-down needs an extreme version of the ‘screening hypothesis’ and would imply that around the world governments are wasting enormous amounts of resources into education without overall pay-off. You now and then get the odd economist arguing it but not without blushing. So if we accept that education is good for you then IQ is also good for you and not having it is bad for you.
So the usual cheap shots at the Bell Curve idea are not sustainable, unfortunately.
A more promising avenue is to see IQ as itself the outcome of a process of investments and that the same investments also cause bigger brains and would help with health, income, etc. This is what Flynn is trying to say. Effectively one can argue that IQ is like being able to drive cars: you need to be rich to have a car, having a car makes you even richer, and it helps if your parents had a car for you to practise on, but this does not imply there is anything genetic about car-driving abilities at the level of racial groups.
Consider all the things you then need to buy into however. For one, you must then argue that those countries with low IQ, brain size, and wealth, are in that position by virtue of some set of historical circumstances that have less afflicted the ones further away from the equator. Plenty of historical stories that give this, so that might be doable but it still needs to be proven. Second, you must then argue that there are strong lock-in effects that make a low IQ-Wealth-Brain equilibrium very stable so that a few generations of wealth increase do not suffice to escape. Third, you must buy into the notion that the individuals who migrated (as slaves or otherwise) were so disadvantaged by their initial low IQ that neither they nor the ensuing generations really managed to escape the relative disadvantage.
Consider this last point more carefully: to rationalise that even for communities that have lived in other countries for generations IQ is still lower and that IQ does correlate strongly between the generations, one must buy into the notion that kids will already very early on have acquired some of the lower IQ from their parents. Indeed, considering the high degree of correlation between separated twins, one must buy into the notion that the womb itself is a place where a lot of the disadvantages of the mother get conveyed onto the children before they are born.
There are mechanisms that fit this kind of hypothesis, including the notion that a wealthy and well-fed mother makes a better womb. Also there is the idea that genes can be turned ‘on’ and ‘off’ depending on circumstances and that genes get passed on set in an ‘on’ or ‘off’ stage. The finding that identical twins are much more similar than non-identical twins could then be due to a shared epigenetics: the identical twins’ on-off switches are synchronised in the womb, whilst this does not hold for the on-off switches of the non-identical twins for whom the same womb produces a different epigenetic effect. I have no idea as to the nitty-gritty of this, but can see how it might happen in principle.
If this is true then initial disadvantages between ethnic groups would then still get washed out as accidental ‘on-off’ switches occur each generation, but would disappear over the course of centuries. For any two or three generations you would then still see a lot of persistence in low IQ and its associated outcomes, but no true long-run genetic causality. To buy into this kind of story means one rejects the notion that heritability inevitably means genetic causality, something that would mean large changes in how genetic effects are inferred throughout the social sciences.
A great example of just the kind of mechanism we would be thinking of is given by height, which is quite heritable. People nowadays are much taller on average than they were in centuries gone past and the amazing thing is that the increase has been quite gradual for several generations, even though a lot of the most obvious ‘benefits of civilisation’ were achieved early on. Thus, at the start of the European height boom in the 19th century, people were perhaps a feet smaller than now but you saw the advent of better hygiene and the demise of infectious diseases. So the first generation was a bit taller. With better nutrition, less hard work and less other sick people to infect the general population, the next generation again was taller. This meant better wombs and richer parents for yet another generation who were fed and looked after even better. We are now talking about the 1930s when families were still huge and hence one couldn’t argue that life was easy yet. Yet, 3 generations on from that one and the Europeans are still getting taller, even in places where they have been traditionally tall already, such as the Netherlands and Scandinavia.
The story of height is hence very much one of an obvious co-movement of increases in health, wealth, and inputs into the next generations. And you can still see very short and very white people in Northern India where the genetic background to the Indo-Europeans is similar but wealth and health are still at the bottom of the scales, which translates into low IQs (and small brains).
The implications of this thinking go further though than just buying into the notion of quite persistent effects that take generations to wash out: it leads to the thought that there is (almost) nothing genetic about IQ at all, even within societies.
That kind of radical counterpoint comes with its own questions though. If there is no long-run genetic marker for intelligence, where does IQ variation within societies come from?
Some would come from the economic position of the parents, ensuring that wealth and health advantages that might be accidentally obtained in one generations get passed on (well-fed wombs again, but also the ongoing effects of being in a confident and positive environment on the ongoing development of the brain). Yet, if that were the only source of variation then societies and families should diverge as advantages get accumulated. This is not the case, because in general the next generation reverts back to the population mean, showing that there must be a strong other source of variation such that the persistence effects of parents helping their young ones are ineffectual against a long-run regression-to-the-mean force because of this ‘other’ source of variation.
What could this wealth-independent source of variation be? Perhaps it is as simple as the embryo and the young baby being lucky enough to get just enough and not too much of everything. A little bit how young sapling trees get to be tall because of accidental circumstances in the forest (more light available and the demise of competitors), so too could it be true that minute differences in the womb and very early childhood could ‘deliver’ very smart and not-so-smart new humans. And the ‘just enough’ factor matters: too much water or sunshine kills off the sapling just as much as too little and when there is the presence of other trees, streams, and other ‘genetically unforeseeable’ circumstances, there is no way to guarantee that the sapling will be ‘most fit’ for each circumstance, meaning that even the ‘perfect tree gene’ needs to be in the lucky position of having a ‘perfect accident’ of adjacent trees, water, etc. So too could it be with humans and intelligence due to accidents in the womb, the very process of egg fertilisation, and very early on in childhood: if intelligence is the outcome of many finely-tuned circumstances which are not under full control and of which the womb can only ‘make an educated guess’ from an ex-ante genetic point of view then more or less accidental circumstances will lead to a large variation in the intelligence of the young kid.
So, if you want, like I do, to hope that IQ and race are truly not causally linked via a genetic and thus somewhat unalterable mechanism, then you will find yourself nudging towards accepting a quite strong version of class-theory (many current disadvantages are perpetuated by the investments of the previous generation), the rejection of heritability as an indicator of genetic causality, willing to accept that there is almost no genetic cause of IQ even within populations, and wondering what the complex underlying source of variation in intelligence really is. That kind of thinking puts you outside of the mindset of most of the biologists and medics working on this, who are wedded to the genetic paradigm and who thus have only a minimalist sophistry standing between them and the Bell Curve …
There has to be some genetic component of IQ. There also has to be an environmental component. Both of these components are subject to a degree of inheritance – the environment is to a large degree shaped by the parents, their relatives, and associates, and genes are inherited biologically.
Because of this, it is going to be very difficult to work out if the differences between various race’s average IQ test results are due to genetics or environmental factors.
I agree that many current disadvantages are perpetuated by the investments of the previous generation. But even with this explanation, there is still room for a racial component (which I believe is very week or perhaps non-existent based on how I see people of different races behave). Short of raising groups of different races under controlled environments, I can’t see how this can be resolved.
LikeLike
Why does there HAVE TO BE a significant genetic component to IQ? It is not clear to me at all that this has to be the case.
One would need to keep up those controlled environments for several generations to get definitive answers from experimentation. I expect we will make progress on some of the underlying mechanisms without that kind of experimentation.
LikeLike
There doesn’t have to be a significant or racially varying genetic component, but there has to be a genetic component because our genetics contains the instructions to make our brains. I didn’t use the word significant, and described uncertainty about how significant it is.
LikeLike
If you want to reject heritability as an indicator of genetic causality, you’d also have to reject relatively new evidence tying heritability to SNP data. On the IQ front:
http://www.nature.com/doifinder/10.1038/mp.2011.85
LikeLike
Hi Jason,
Yes, it’s a big mountain to climb, I agree. What gives me hope is the spectacular increases in IQ that Flynn reports for countries that increase their education for a couple of generations. That, plus a lack of clear smoking guns on the IQ front, ie it is not clear which ‘smart gene’ is supposedly lacking in particular ethnic groups, so there remains a mountain to climb for the genetic brigade too.
LikeLike
Have you considered the Flynn-Dickens model? Under that model, you don’t have to reject the within population heritability and it saves you from looking to epigenetics.
http://dx.doi.org/10.1037/0033-295X.108.2.346
LikeLike
Yes, but I find it a bit convoluted (with sufficiently complex unproven interactions between environments and genes, almost any outcome becomes possible). Also, I am somewhat convinced by these ‘womb’ stories, ie the idea that a lot of important variation is not genetic but related to circumstances in the womb. As soon as one allows for that possibility, one of course already has multi-generational feedback loops. It’s the particulars of how that womb effect can be persistent and particular enough to rationalise the various regularities that seems to me the main puzzle. I am hoping it negates the role for true genetic differences in intelligence, but as a mere bystander in this area, we will just have to wait and see….
LikeLike
ps. Thanks for the link to the article. It would take me at least a week to properly unpack this thing, but a casual reading doesn’t fill me with confidence about their interpretations.
For instance, they say “Finally, using just SNP data we predicted ~1% of the variance of crystallized and fluid cognitive phenotypes in an independent sample (P=0.009 and 0.028, respectively).” which means their ‘single nucleotide polymorphisms’ explain about zip of intelligence out of sample.
Their numbers on genetic variation hence once again derive almost entirely off the similarity in intelligence between people who grew up close together and who thus share a lot of genes and SNPs. That could thus incorporate all kinds of effects, ranging from neighbourhood effects to family effects and even genetic effects.
The article exemplifies though that there are lots of tricks in the basement of this literature. Take this paragraph: “For gf in the Manchester and Newcastle samples, empirical Bayes’s estimates for each individual were obtained from a random effects model fitted by maximum likelihood to the standardized age-regressed residuals obtained for each sex from the Alice Heim 4 test and the Cattell Culture Fair test scores. All of the phenotypes were corrected for age and sex (with the exception of Manchester and Newcastle gf, which was derived separately for males and females) and the standardized residuals were used for all subsequent analyses.”
In two sentences they here quickly mention all kinds of corrections, standardisations, and methods that need a lot of detailed background assumptions (such as prior for their Bayesian methods, an age-window for their residuals, and some kind of comparability assumption between samples). In economics, this would take one 40 pages of agonising and defending, and these guys just quickly mention it. All kinds of things could thus be rationalising their results. Unless I really spend lots of time getting on top of how they handle their data, code their tests, etc., I wont be able to judge whether this article really establishes anything new at all. Can you?
LikeLike
The article’s does show something new – that 40% to 51% of variation in intelligence was due to genetic variation contained in the SNP sample. Sure, the specific gene’s aren’t pinpointed, but they are there.
The statistics aren’t quite as reliant on ‘tricks’ as you imply. The paragraph you quoted was an attempt to estimate fluid ‘g’ from some intelligence tests, which is not central to the headline finding. If you look at the paper that kicked this type of analysis off (on height – http://dx.doi.org/10.1038/ng.608, with a useful commentary here – http://dx.doi.org/10.1375/twin.13.6.517), the method does not appear as convoluted. And here’s an example of economists using this technique – dx.doi.org/10.1073/pnas.1120666109
I’m open to being convinced this technique doesn’t work or has flaws, but from what I have read (as a mere bystander) it seems solid. And as aside, Peter Visscher from UQ is behind a lot of this work.
On your earlier comment on womb stories, I haven’t seen any mechanism (proposed or demonstrated) to explain why the womb environment would be different for fraternal and identical twins to an extent that could explain the heritability findings. Was there a piece of evidence that tipped you over on that point? I’m open to the potential for short-term (couple of generation) effects in the womb environment, but these epigenetic (or whatever they may be) effects don’t necessarily undermine heritability findings. They could even be causing us to underestimate the effect of fixed biological factors on our phenotypes.
LikeLike
Their numbers on genetic variation hence once again derive almost entirely off the similarity in intelligence between people who grew up close together and who thus share a lot of genes and SNPs. That could thus incorporate all kinds of effects, ranging from neighbourhood effects to family effects and even genetic effects.
They excluded all pairs of individuals whose genomic relationship was equal to or closer than that of fourth cousins. Thus no family effects. The fact that they couldn’t find single SNPs that explain more than a tiny bit of the variance is a statistical power issue, but it doesn’t affect the conclusion that considered together, the SNPs explain a lot of variance. Because the subjects were unrelated and did not share family environments, the correlation between their genetic and phenotypic similarity must indicate causality. Note that they also controlled for population structure using PCA.
Partialling out age and sex variance is a standard procedure in IQ studies, and I don’t see how that could bias their results in favor of a genetic model. The assumption made is that there are no sex- or age-specific genetic interactions that affect IQ variance. If there in fact were such interactions, they would attenuate the estimate of heritability. The same applies to the question of comparability of different samples. The empirical Bayes method is used to get more accurate estimates of factor scores, and it’s difficult to see how that could bias the results, either.
The results of Davies et al. have been replicated in other samples. See here, here, and here, for example.
If you disagree with the GCTA results, please explain how they could be obtained if the conclusion of genetic effects is in fact wrong. If you cannot provide even a hypothetical alternative model, you don’t have a case. I have discussed this research with others that dispute them, but none of them have been able to provide a plausible alternative model. Vague hand-waving about possible confounds doesn’t cut it.
LikeLike
This post makes the hereditarian case for us. At the moment, I don’t really see much to add on top of what the other commenters have.
I will say this: this seems to be running Occam’s Razor in reverse. One should ask oneself which explanation for all this evidence has the fewest terms and requires the fewest assumptions?
Second, it’s generally wise (not to mention good scientific practice) to adjust one’s beliefs on the basis of the evidence, not to try to get the evidence to fit a particular a priori belief.
Well I will comment on one thing for completeness’s sake:
This part is complete nonsense. See:
Epigenetics smackdown at the Guardian « Why Evolution Is True
LikeLike
Typical bogon reply. Quick link to an article only related to a much stronger version of what I am proposing, a repeat of the fact that I wish to believe something (not actually believe something) and then a haughty couple of ad nominems. Thanks for visiting. It’s been great.
LikeLike
I have no idea how gather that my post contains an ad hominem, but OK.
My link (one of many on the topic I could have chosen) doesn’t just address Lamarckian epigenetics but addresses within generation change (which is also overblown).
In any case, turning to within-utero epigenetics to explain the higher correlation between MZ vs DZ twins is an Occam’s Butterknife explanation. We know that there is extra genetic similarity between MZ twins, but you’re proposing adding an extra element, one that isn’t necessary given the evidence we have.
Note, my comments don’t imply that you necessarily believe these claims. One can entertain and evaluate the merits of a claim without necessarily believing in it oneself.
LikeLike
It’s a mistake to treat IQ differences between generations, countries, and races as qualitatively the same. There are several studies demonstrating that cohort differences in IQ reflect test bias and therefore cannot be adduced as evidence for increases in intelligence. The same goes for differences between countries. This is the locus classicus for the former, while this shows that even the PISA tests, which were specifically devised for the purpose of international comparisons (unlike IQ tests), generally lack international comparability. Just about the only prominent population difference that can be treated as a genuine ability difference is the black-white IQ gap in America (see discussion in in the first link above).
To make group comparisons when measurement invariance has not been established is often tantamount to committing a category error. It’s like comparing the heights of individuals when they have been measured using different units of length.
The truth that IQ is genetically heritable is not contingent on twin studies. Other research such as those making use of adoptees and DNA-based similarity indices affirm the results of twin studies.
The epigenetics stuff you describe is completely fanciful, and refuted by, for example, the sort of genomic studies Jason Collins refers to.
Even if IQ is heritable within all populations and fully reflects ability differences between those populations, it’s of course still possible that population differences in IQ have an environmental etiology. It just requires that the lower-scoring group be severely disadvantaged in trait-relevant environmental circumstances compared to the higher-scoring group. The higher the heritability is within populations, the larger the environmental discrepancy must be for a non-genetic explanation of the group difference to be viable.
For example, if the IQ gap is 1 SD (15 points), environmentality (the complement of heritability) is 25 percent, and the gap is assumed to be entirely non-genetic, then the trait-relevant environment of the average member of the lower-scoring group must be worse than that of 98 percent of the members of the higher-scoring group (assuming that environmental effects are more or less normally distributed). However, if the average environment of the lower-scoring group is superior to that of the most deprived (bottom 2 percent) members of the higher-scoring group, then at least some of the gap must be genetic in origin.
LikeLike
Again someone eager to dismiss anything that starts with epigenetics! There is the distinct smell of in crowd conservatism about this. Are you a producer in this literature and can you truly hand-on-heart tell me the current crop of suggestive (but indeed not conclusive) studies appealing to epigenetics is basically mirage-hunting by soon to be sidelined scientists? I will wait and see.
This ‘we know its genetics’ line presumes you know the genetic mechanism associated with intelligence. Well, what is it? I can’t find it in the literature so enlighten me. People infer its genetics because of a lack of alternative simple stories that rationalises the same data, but that is second best. Give me the recipe.
I must say that I find the combination of ‘its genetics’ plus ‘one can’t compare intelligence’ humorous. Like watching someone contort themselves in many bends to have their cake and eat it. If its genetics, then what miracle prevents one from saying that groups without the ‘right’ genes lack a type/level of intelligence that the group with the right genes does have? One is left with a thin veneer of sophistry. So you should probably hope for the same thing I do!
LikeLike
Are you a producer in this literature and can you truly hand-on-heart tell me the current crop of suggestive (but indeed not conclusive) studies appealing to epigenetics is basically mirage-hunting by soon to be sidelined scientists?
The environment of course influences how genes are expressed, but epigenetics in this sense is captured in the non-genetic variance components in heritability studies. It’s not confounded with heritability.
If you’re talking about transgenerational epigenetics, that’s where the fanciful comes in. I don’t believe that epigenetic effects on DNA sequence variation could explain the heritability of IQ. This is simply because of a lack of evidence for such effects, and because the available evidence on IQ heritability is incompatible with epigenetic explanations. I think the heritability of IQ is explained by a mainly additive polygenic model because the data fit it well and the model has withstood numerous attempts at falsification.
If you want to dispute the standard polygenic model, you must show that your preferred model fits the data equally well or better. For example, how does your model explain the results of GCTA studies which show that genetic similarity among unrelated people predicts IQ similarity between them?
I must say that I find the combination of ‘its genetics’ plus ‘one can’t compare intelligence’ humorous. Like watching someone contort themselves in many bends to have their cake and eat it. If its genetics, then what miracle prevents one from saying that groups without the ‘right’ genes lack a type/level of intelligence that the group with the right genes does have?
The reason I say it is because it’s true. If people born 100+ years ago were tested with today’s IQ tests, a very large fraction of them would score in what is the retarded range by our standards. Yet they didn’t behave like retarded people but lived normal, productive lives. Clearly the measurement instrument is invalid in this case, and statistical analyses of measurement invariance indicate that that’s indeed the case.
I believe that there are heritable IQ differences between populations, because, given the highly polygenic genetic architecture of intelligence, identical distributions of IQ genes in all populations is an extremely unlikely state of affairs. However, you obviously cannot prove that a phenotypic difference in IQ reflects genetic differences just by pointing at that phenotypic difference, for reasons I’ve elucidated above.
LikeLike
I posted a rather detailed comment on this. Why wasn’t it published?
LikeLike
“Suppose you wanted to believe, as I do, that intelligence and vague ‘racial groups’ are, on the whole, unrelated from a long-run perspective.”
I’m not being sarcastic when I ask: What do you mean by “unrelated”?
Men are taller than women, on average. Is height unrelated to gender?
I’m not terribly concerned about the causes of IQ. It clearly exists. It clearly has distinct racial distribution patterns, with “race” being self-reported. I am perfectly capable of believing all that without ever once thinking that IQ is linked to race in any way, just as I don’t think my gender causes my height. It just makes certain outcomes more likely.
LikeLike
An interesting article. But the fundamental premise is that IQ tests are a primary measurement system. In fact they are not, as clearly demonstrated by Flynn’s work. Consequently, comparison of IQ test data across culturally disparate groupings is meaningless.
LikeLike
I think economists neglect this topic at their peril. I believe there is a (small) genetic component to differences in IQ across different populations. But most of the differences (in both IQ and economic performance) are explained by environmental factors such as the level of dignity/ freedom/ nutrition.
I’ve elaborated these models at some length in a booklet here: http://sanjeev.sabhlokcity.com/Misc/Notes-on-IQ.doc, and in numerous blog posts. I also demonstrate that without removal of the caste system in India, the “national IQ” of India will not increase substantially.
LikeLiked by 1 person
P, Jason,
You guys have clearly read the linked SNP+genes article, and articles in the same vein, much more carefully than I did in the hour I gave it. Still, your replies really don’t warm me to it.
To start with the most important thing, what is the variation on which their results are based? Essentially the cross-correlation between a large amount of shared genes/SNP and IQ-type tests. Just ask yourself, who shares the same genes and SNP? Well, people living close to each other, of course. People in the same family, village, county,etc. Even if you kick out recorded family pairs up to 4th cousins, you are still identifying the whole thing off genetic and SNP proximity and will thus still be looking at villages and undocumented family ties. It makes me think of my father whose family lived for 500 years in the same village. Everyone was family after that time, simply not fully recorded.
So the variation these guys work with will be completely dominated by spatial proximity. This gives one all kinds of alternative stories: shared schools, particular teachers, local economy, local environment, etc. It is up to the authors to convince us that their results have no influences in them of unobserved shared circumstances, not me to crawl over their data and re-do it whilst adding in other data on those areas. I leave that to ambitious young scientists who need to make a name for themselves in that field.
It is also very hard to see how one can escape this trap: of course genetic and SNP proximity will overlap very closely with spatial proximity, family lines, social groups, etc. One would have to randomly allocate twins to families all over the world to escape this trap. But it does mean that these highly artificial models, with their many unobserved latent constructs and smoothing parameters (how important are those priors?), are ultimately not all that convincing of anything. Until they of course do find genetic smoking intelligence guns and actual replicable mechanisms. Till that time, the field is still open and I can dream about epigenetic intergenerational mechanisms that rationalise the data without being stuck in a world where intelligence is a matter of selective breeding.
I must say, P, that you are being inconsistent with respect to your interpretation and use of these IQ tests. What do you think the linked study that you seem to like uses but very similar tests? And if you think there are interpretation issues over time and across culture, why shouldn’t this hold across villages in the UK? Anyhow, all that is already covered in the post.
Btw, Jason, the womb stories I was thinking of were not directly on different types of twins, but more on the effects of how the mother is stressed or ill on the unborn child, with knock-on effects for the further generations (who, in egg terms, are also then being born). On a related vein, I just saw an article in the New Scientist (I think) speculating about intelligence as being subject to a highly random process, much like I hypothesised in the post.
LikeLike
To start with the most important thing, what is the variation on which their results are based? Essentially the cross-correlation between a large amount of shared genes/SNP and IQ-type tests. Just ask yourself, who shares the same genes and SNP? Well, people living close to each other, of course. People in the same family, village, county,etc. Even if you kick out recorded family pairs up to 4th cousins, you are still identifying the whole thing off genetic and SNP proximity and will thus still be looking at villages and undocumented family ties. It makes me think of my father whose family lived for 500 years in the same village. Everyone was family after that time, simply not fully recorded.
As I explained above, in these studies, determination of “family” relationships is not based on recorded relationships, but on genomic measures. All pairs that are genetically as or more similar than fourth cousins are excluded. Thus the sort of clustering by geography that you describe is not possible.
Additionally, in many of these studies the subjects are drawn from a large geographic area. For example, the Swedish study I referred to above is based on a sample of almost all male twins born in Sweden in certain years. One twin from each pair was selected, and in the resulting sample all pairs whose genomic similarity corresponded to fourth cousins or closer were excluded. Because twinning is a random process, the Swedish sample is nationally representative, and most people in it have probably never met each other (the IQ scores were obtained upon induction into compulsory military service).
It is up to the authors to convince us that their results have no influences in them of unobserved shared circumstances, not me to crawl over their data and re-do it whilst adding in other data on those areas. I leave that to ambitious young scientists who need to make a name for themselves in that field.
Your criticisms of these studies have so far consisted of non sequiturs that have no bearing on the validity of their results. Your hope that ambitious young scientists will somehow refute this research is not going to be fulfilled. The GCTA method has been around for several years now, and the response from the traditional critics of behavioral genetics has been instructive: They are completely stumped. These new methods vindicate the classical methods of behavioral genetics (such as twin studies), and show that the voluminious literature denying the basic results of behavioral genetics is without merit.
I must say, P, that you are being inconsistent with respect to your interpretation and use of these IQ tests. What do you think the linked study that you seem to like uses but very similar tests? And if you think there are interpretation issues over time and across culture, why shouldn’t this hold across villages in the UK?
IQ tests are a valid measure of individual differences when the individuals in question share the same culture. Tests are generally not valid when the subjects don’t share a culture, e.g., different generations of people from the same country. Therefore, individual differences within a group may be valid, reflecting both genetic and environmental influences, while between-group differences aren’t because they reflect just test bias.
However, assuming that the genetic architecture of intelligence is the same across generations, the GCTA method can be used to assess heritability in samples drawn from different generations provided that the test scores are residualized for age. (If the genetic architecture differs between generations, GCTA will of course produce low heritability estimates.)
LikeLike
Nope, being representative and using just one of the twins doesn’t kill off the proximity possibility at all: the variation one is relying on is still genomic similarity which will be strongly correlated with spatial proximity.
It’s not a hard point to get. Think of the following analogy: suppose you are trying to get at the correlation between skin colour and wearing hats against the sun. You start out with a representative sample of the human population. Skin colour of course gets lighter the further away you go from the equator, together with the prevalence of sun-protecting hats (making this up, but you get the point, I hope). So you conclude that darker skin colour explains 51% of sun hat wearing.
It really doesn’t matter if you in this analogy start out with a representative sample. What does matter is whether you control for environmental and regional factors.
I am amused to see you think of culture as (0,1), ie no differences within a country and insurmountable differences in comparability across them. Convenient sophistry, I am afraid. The post however really already covers this point.
Oh, and if you want me to respond to further comments, please use your actual name. I so detest cowards who hide behind pseudonyms and then make snide remarks.
LikeLike