How the race, intelligence, and genetics question will semi-resolve within the next 10 years

By Razib Khan | May 28, 2013 12:55 am

Prompted by my post Ta-Nehisi Coates reached out to Neil Risch for clarification on the nature (or lack thereof) of human races. All for the good. The interview is wide ranging, and I recommend you check it out. Read the comments too! Very enlightening (take that however you want).

When it comes to this debate I have focused on the issue of population substructure, or race. The reason is simple. Due to Lewontin’s Fallacy it is widely understood among the “well informed general public” that “biology has disproved race.” Actually, this is a disputable assertion. For a non-crank evolutionary biologist who is willing to defend the race concept for humans, see Jerry Coyne. When you move away from the term “race,” then you obtain even more support from biologists for the proposition that population structure matters. For example, a paper in PLoS GENETICS which came out last week: Analysis of the Genetic Basis of Disease in the Context of Worldwide Human Relationships and Migration. In other words, it is useful to understand the genetic relationships of populations, and individual population identity, because traits correlate with population history. Barring total omniscience population history will always probably matter to some extent, because population history influences suites of traits. If nothing in evolutionary biology makes sense except in light of phylogeny, much of human biology is illuminated by phylogeny.

But that doesn’t speak to the real third rail, intelligence. Very few people are offended by the idea of the correlation between lactase persistence and particular populations. Neil Risch says in the interview with Coates:

One last question. Your paper on assessing genetic contributions to phenotype, seemed skeptical that we would ever tease out a group-wide genetic component when looking at things like cognitive skills or personality disposition. Am I reading that right? Are “intelligence” and “disposition” just too complicated?

Joanna Mountain and I tried to explain this in our Nature Genetics paper on group differences.  It is very challenging to assign causes to group differences. As far as genetics goes, if you have identified a particular gene which clearly influences a trait, and the frequency of that gene differs between populations, that would be pretty good evidence. But traits like “intelligence” or other behaviors (at least in the normal range), to the extent they are genetic, are “polygenic.” That means no single genes have large effects — there are many genes involved, each with a very small effect. Such gene effects are difficult if not impossible to find. The problem in assessing group differences is the confounding between genetic and social/cultural factors. If you had individuals who are genetically one thing but socially another, you might be able to tease it apart, but that is generally not the case.

In our paper, we tried to show that a trait can appear to have high “genetic heritability” in any particular population, but the explanation for a group difference for that trait could be either entirely genetic or entirely environmental or some combination in between.

So, in my view, at this point, any comment about the etiology of group differences, for “intelligence” or anything else, in the absence of specific identified genes (or environmental factors, for that matter), is speculation.

In response to this commenter Biologist states (note, I know who this is, and they are a biologist!):

Risch writes: “…the explanation for a group difference for that trait could be either entirely genetic or entirely environmental or some combination in between. … So, in my view, at this point, any comment about the etiology of group differences, for “intelligence” or anything else, in the absence of specific identified genes (or environmental factors, for that matter), is speculation.”

This is essentially correct. The quality of available evidence on which to estimate the contribution of genetic versus environmental factors to group differences in cognitive ability scores is quite poor by biomedical research standards — maybe more in line with standards for social science (I’m only half joking).

In light of that, one is forced fall back on to one’s priors. Without trying to speak for Risch, it is generally considered appropriate to adopt a uniform prior in the absence of other evidence. Under a uniform prior, “…the explanation for a group difference for that trait could be either entirely genetic or entirely environmental or some combination in between.” Maybe Risch would propose a different prior.

In fact, the uniform prior says that there’s a 25% chance that the explanation is 0% to 25% genetic, a 50% change that the explanation is 25% to 75% genetic, and a 25% chance that the explanation is 75% to 100% genetic. Obviously many people who write about this topic do not adopt a uniform prior. [my emphasis -Razib]

As Risch observes above intelligence is highly polygenic. There’s a fair amount of genomic evidence for this now. In other words the likelihood is not high that we will be able to account for the differential distribution of IQ between any two populations by differences in allele frequencies. Even if we do find the allelic differences, they’ll account for far too little of the variation in the trait. But there is another way we can get at the issues. Others have pointed out exactly how we can get more clarity on the race and IQ question before, so I’m not being original. And since I suspect that within the next decade this sort of analysis will likely be performed at some point somewhere because the methods are so simple, I might as well be explicit about it.

Let’s focus on the black-white case in the American context. On intelligence tests the average black American scores a bit less than 1 standard deviation below the average white American. As I’ve observed before the average black American is ~20% European, but there is variation around this value. Because the admixture is relatively recent (median ~150 years before the present) there is a wide range across the population of ancestry. In fact, the admixture is recent enough that siblings may even differ in the amount of European ancestry on a genomic level. An additional issue which is of relevance is that the correlation between ancestry and physical appearance in mixed populations is modest. By this, I mean that there are many individuals who are more European in ancestry in the African American population who have darker skins and more African features than those who have less European ancestry. Obviously on average more European ancestry predicts a more European appearance, but this is true only on average. There are many exceptions to this trend.

At this point many of you should have anticipated where I’m going. If the gap between blacks and whites on psychometric tests is totally driven by genetic differences between Africans and Europeans, then the gap should be obvious between pools of individuals of varying levels of European ancestry within the African American population. It seems unlikely that it would be that simple (i.e., all driven by genes without any sensitivity to environmental inputs or context). Therefore I suspect some design where you compare siblings would be more informative.

In a model where all of the between group differences are due to environmental inputs then genomic ancestry by geography within family should add little in terms of prediction of the phenotype. More plainly, when accounting for other variables which might correlate with ancestry (e.g., skin color), how African or European a sibling is should not influence outcomes on psychometric tests when looking at large cohorts of sibling pairs if those differences track nothing more than social construction/perception of race. If on the other hand there are many alleles of small effect distributed throughout the genome correlated with geographic ancestry which affect the final phenotype then adding ancestry as an independent variable into the model should be informative. This sort of indirect inference has already been performed with a character similar in genetic architecture to intelligence: height. Researchers have found that African Pygmies with more non-Pygmy ancestry are taller.

Ultimately I say that this issue might semi-resolve, because I think a hereditarian position in terms of group differences is not going to be tenable if the correlations with ancestry do not run in the direction expected within admixed populations. This sort of model is relatively straightforward in its predictions, and appeals to parsimony. Trying to salvage it with non-additive genetic variance is going to complicate matters. In contrast, those who champion the opposite position often dispute the very characterization of intelligence as a trait in the first place, so I presume that they would still exhibit skepticism if there was a correlation between genomic ancestry and the trait.

Addendum: I want to be clear: with the widespread availability of data sets and crappy security of said data sets this analysis is probably a few SQL joins away in 10 years.

CATEGORIZED UNDER: Genetics, Psychology
MORE ABOUT: I.Q., Intelligence, Race

Comments (98)

  1. I was thinking a precursor to this type of study had already been done. Am I wrong? I want to say it was something like blacks adopted by white parents or something along those lines. I guess we could just have Brad Pitt test all of his kids to find out – it’s definitely a large enough sample size:)

    • razibkhan

      not to my knowledge. though perhaps you heard about something in press? to be clear the key difference is to genomically ascertain ancestry on a quantitative level, rather than just binning racial category.

      • I can’t remember anything specific so I thought I would just put that out there. I might be combining several different memories of various papers with Steve Hsu’s suggestion that you could eventually be able to predict someone’s IQ just by scanning their genome.

  2. marcel proust

    I’m unusually slow this morning, and not certain that I understand the 2nd to last paragraph (“In a model where all ….”)

    i think you are saying the following.

    1) We want to test the hypothesis that a genetic basis exists for the poorer performance of African-Americans on intelligence tests.*

    2) Run a regression (or comparable statistical analysis) on a sample of individuals, with intelligence-test-performance as the dependent variable, and every genetically determined trait that we know differs between (sub-Saharan) Africans and non-Africans as independent variables.

    a) Include those traits that are clearly visible and are very unlikely to influence intelligence to capture anything related to the social construction of race.

    b) Include any other (childhood?) socio-economic/demographic traits as well to capture any environmental factors related to performance on intelligence tests. These are likely to wash out among siblings, other than birth order and gender (distinct from sex and perhaps not genetically determined).

    3) Include a measure of degree of African ancestry. If this … not sure here: is statistically significant? substantially reduces the RMSE? then what it is capturing is the many genetic differences between those with African ancestry and others that influence intelligence-test-performance and are individually of small effect.

    Is this a correct interpretation of that paragraph?


    *I chose this wording to avoid any discussion of the relation between intelligence and IQ scores. It is not directly relevant to my question.

    • razibkhan

      basically. until recently we couldn’t directly test % of ancestry x across siblings with any precision. now we can. if intelligence variation is controlled by variants at many genes across the genome correlated with ancestry, then this ancestry, which is to a great extent socially cryptic (e.g., a 45% vs. 55% difference is not usually visible to a high degree of accuracy on the individual level), should have an effect.

      to be clear i’m particularly interested in cases where the sibling phenotype goes in the opposite direction from their ancestry. this could test the model that social feedback is driving the outcome (or not).

  3. toto

    “I think a hereditarian position in terms of group differences is not
    going to be tenable if the correlations with ancestry do not run in the
    direction expected within admixed populations.”

    Well… even then, some people could point out that the Europeans who provided the admixture were not necessarily a random sample of all Europeans.

    If you accept the possibility of cross-racial assortive mating, die-hard hereditarians could argue that maybe the whites who provided admixture had a surplus of “dumb” genes, masking the totally-for-realz between-group-variance.

    Note that it could also be invoked in the other direction, if there is a possibility that the admixture comes largely from wealthy (supposedly “smart”) slaveowners and their families – you could be mistaking the within-(white)-group variance for between-group variance.

    • razibkhan

      yes, and it starts to get more complicated. though a ~1 sd difference is quite large. skeptical that assorting could be so extreme.

      • Emil Kirkegaard

        Remember regression to the mean would reduce the strength of assortative mating.

        • I’ve never understood the rationale around how “regression toward the mean” works.

          I mean, I understand in theory that people who are especially bright (or especially dumb) could be so due to a few genes. While they can pass these genes on to offspring, the chances are high their partner will have genes which “turn off” their own genes regarding high IQ in some way, which diminishes (on average) their kids intelligence.

          But I can’t figure out why (as Steve Sailer has suggested in the past) the regression to the mean should be based upon the average for you race, not the population at large. Thus if you have three sets of parents, each with a 120 IQs, but they are black, white and Asian, the black kids will score lower, the whites middling, and the Asians higher.

          Given intelligence seems due to hundreds of genes, you’d need to posit that there were special characteristics to the genomes of different population groups as a a whole. Even if you argued the 120 IQ Africans “rolled sixes” and thus have a very rare genetic combination, it’s difficult to argue that their IQ should regress more than other groups, unless Africans have an excess of genes which actively lower IQ (which, absent active selection for low intelligence, would be weird).

          Indeed, all things considered, if children really do regress to the mean IQ of their racial group, and not the human race as a whole, it would seem to be more indicative of social experiences shaping IQ to some degree.

          • Emil Kirkegaard

            I have no firm idea of the subject. Perhaps Razib can post about it explaining how it work, biologically? Would be preciated! 🙂

            Richard Lynn discusses it in his 2001 book:
            “The essential error in the arguments of Scarr, Eysenck, Preston and Campbell, and others who have argued that regression to the mean would negate eugenic programs lies in the assumption that regression to the mean of the descendants of extreme groups continues for a number of generations until they reach the mean of the population. Contrary to this assumption, regression only occurs in the first generation of children. In a selective breed- ing program, the children of the selected parents establish a new mean, which is maintained in subsequent generations. The genetics of this is explained by D. S. Falconer (1960) in his Introduction to Quantitative Genetics, where he states that in selective breeding programs, “provided there is no other reason for the gene frequency to change, the population mean will be the same in the generations following as in the F2,” that is, the first generation (p. 259). This has been explained more recently by Loehlin (1998), who points out that this is the error in the thesis of Preston and Campbell (1993), that dys- genic fertility has no long-term effect because although the child generation has a lower IQ than the parental generation, subsequent generations gradually regress back to the original mean. ”

            Razib said that he has read the book in question, so surely he must know the details. 🙂

          • razibkhan

            falconer quotes are good. the simplest way to understand regression to the mean in quantitative genetics is this.

            phenotype in a population is controlled by genes & environment. in a complex/quantitative trait (height & IQ) the heritability is a measure of the proportion of variation which is controlled by by genetic variation in the population. so for the following we are talking about populations only.

            imagine then that you select from a population which is >2 standard deviation above the norm in height. you know that the individuals are enriched for alleles which drive greater height, because height is highly heritable. but there is an ‘environmental’ component of height. when you are selecting ppl for deviation from the median and pair them up the genetic part is pretty much guaranteed, because those individuals carry those genes. not only that, but you know that the offspring are going to have a 50/50 mix of those genes (height is distributed across so many genes that should work to a first approximation). you don’t have any good model for what ‘environmental’ factor that the parents had which might have given them an extra post. it could just be random developmental crap which is totally uncontrollable. so depending on how much of the variance is environment, there is going to be some regression back to the mean, because the offspring aren’t likely to have inherited that fluke benefit/cost.

            think of the ‘environment’ in quant. gen. as ‘noise’ which basically messes up with your prediction equation. the more environment a quantitative trait has controlling its outcome, the less you can reasonably select for individuals who are going to produce similar offspring. and this is why heritability can be thought of as a measure of the correlation of the traits of the parents and offspring.

            this is also why regression doesn’t go back to some magical population average. it’s not about populations, but about the information you have. once there is one generation of regression, then it’s done, because the noise effect remains the same over the generations.

            hope that’s clear. you can also think of it as a bivariate normal distribution with one axis being genetic variation, and the other being the enviro/noise component. selection operates only along the genetic axis.

          • Thanks Razib, that makes perfect sense.

            One quibble I have though. In some cases, environmental effects on tall people will not only have boosted their height, but limited it as well. Therefore, you cannot merely use the percent of variance caused by the environment to approximate the regression to the mean. But I’m sure this is accounted for in some ways by calculations of regression to the mean.

            Regarding intelligence in particular, it’s hard to see how environmental regression toward the mean works however. Someone who happens to have a higher IQ than their parents is highly likely to marry a smarter spouse, and raise their children within a smarter peer group than they themselves grew up in. This is particularly true because studies suggest as you climb up the SES scale, the percent of IQ due to genetics rises.

          • razibkhan

            direction isn’t important. i assume it goes both ways randomly. the key is that it is a noise parameter.

            suggest as you climb up the SES scale, the percent of IQ due to genetics rises.

            heritability doesn’t go above 70 to 80%. so you would still regress 1/5 or back toward the mean for the population median.

            for quantitative traits usually it is hard to squeeze all the ‘environment’ out. that’s because there are random developmental things you can not control. genes are only so powerful 🙂

          • James Thompson

            this is also why regression doesn’t go back to some magical population average. it’s not about populations, but about the information you have. once there is one generation of regression, then it’s done, because the noise effect remains the same over the generations.

            Dear Razib, Thanks for your explanation. I am seeking to understand how the first generation “knows” it is the first generation. Surely regression to the mean exists whenever two variables are less than perfectly correlated, and that is true for all generations, because the correlation between parental and child characteristics are never unity. Generations will regress and continue regressing for ever, and whilst that is happening a population average will be maintained, (and a population standard deviation) as each family in each generation oscillates around the mean for people with those sorts of genetics.

            That is my argument, which I am aware may be wrong, but I am trying to understand, since I regard differential regression to the mean as the strongest argument for a genetic component to racial differences in intelligence.

          • razibkhan

            which I am aware may be wrong

            if your argument is correct than darwinian evolution is wrong.

          • James Thompson

            My apologies then. I know this is tedious, but can you recommend some basic homework for me to help me understand regression to the mean as it relates to intelligence?

          • razibkhan

            let’s try this then. imagine that you have a quantitative trait with a continuous distribution which is heritable. the trait is 50% heritable. you select individuals two standard deviations above the median in trait value.

            these are exceptional individuals. some of this is surely genetic. in fact, you know the magnitude, since heritability is 50%. but the non-heritable component also matters, and contributes their deviation.

            if you have the two sigma dev individuals mate and produce offspring then the resulting population will regress halfway back to the original population mean. why? because genes are heritable, but the non-genetic component is not. so all subsequent generations will not regress to this new median, because that’s the genetic median.

            obviously an interesting fact due to this is that very heritable traits exhibit less regression. that’s because you have a much more “honest” signal of the genetic character in relation to the phenotype.

            i’m having a difficult time parsing your original comment, but this is i think the key:

            Generations will regress and continue regressing for ever, and whilst that is happening a population average will be maintained, (and a population standard deviation) as each family in each generation oscillates around the mean for people with those sorts of genetics.

            selection transforms the genetics. in direct proportion to how heritable those traits are. there isn’t a magic population median which traits will regress back toward.

          • James Thompson

            Thank you very much for your additional explanation. I am in agreement with you on all the main points, so any problems are probably due to my poorly expressed comments earlier. I certainly understand that very heritable traits regress less. For example if intelligence is .8 heritable (as might be indicated by comparing parents with their offspring tested in young adulthood) then there will be less regression. The children of bright parents will only fall back 20% of the distance back to the mean. However, I had assumed that the regression effect occurred throughout the whole intelligence range. For example, if intelligence is .8 heritable parents of mean IQ 130 will have children of mean IQ 124. Parents of IQ 110 will have children of IQ 108 etc. Two questions: 1) Are you saying there is more noisy “luck” at the extremes of intelligence, or is it uniform through the range, as calculated from the overall heritability of the the trait in question 2) Are you also saying (here I get my apology in early, because this was the bit I did not understand) that regression occurs in the first generation only, and not in subsequent generations?
            Hope my questions are clearer than before.

          • razibkhan

            yes, regression occurs across the whole range. what was talking about is that the value the regression regresses too is not fixed. a common misconception is that selection/assortative mating is futile because populations ‘regress to the mean.’ but of course in selection regimes regression see’s the previous generation mean only.

            i’m pretty sure the confusion here was only semantic.

          • James Thompson

            Great. Thanks for your replies.

          • Gregory Cochran

            it works like this. Imagine that we have two parents with IQs of 120, drawn from a population with an average IQ of 100, and for which the narrow-sense heritability of IQ is 0.6 . Then, on average, 60% of those 20 extra points are genetic, while 40% are the product of good environmental luck. The next generation gets those parental genes, but the luck dice are thrown again, and on average yield a zero. So the average IQ for those kids is 114.

            Do this again with a population whose average IQ is 85.They have 35 more IQ points than the average of their population. 60% (21 pts)will be genetic, 40% environmental. So the average IQ of their kids is 106.

            You can think of it this way: scoring a lot higher than your group average takes more luck than scoring moderately above your group’s average – and in the next generation, the good luck goes away.

    • Emil Kirkegaard

      I think such effects are likely (assortative mating), but much smaller than the ~1sd difference between US whites and blacks. But it is testable. One could gather a sample where one has the parent’s IQ as well. Then one knows the parents are representative.

      Also remember regression to the mean, which reduces these effects.

      Besides, as noted in the comments on Razib’s last article on race+IQ. The hereditarian prediction was confirmed for SES, education level, income, and occupation. These are all well known correlates of intelligence, all of them going in the predicted direction with a huge sample. Only IQ and g needs to be confirmed too for this style of study.

      After that, the only thing left to confirm is to identify specific genes for g, and check their frequency in different races. That’s the decisive test, no? After that there can be no more arguing. Even now the environment-only position is stupid, but it will be worse when it fails the predictions on these studies described.


      Table S2.

  4. genobollocks

    And then you still wouldn’t know about whether any differences are due to gene-environment interactions, ie. one group does worse in the American environment (additionally, for less obvious reasons than discrimination).
    You could also look at the historical probability of mixed marriage given what we know on assortative mating for intelligence and intelligence-openness correlations.

    So, where you end up after receiving more fairly indeterminate information may still depend on your prior.

    Of course, most educated people don’t adopt a uniform prior, because we have good _experimental_ evidence for the negative effects of discrimination from stereotype threat research.

    Experiments are stronger evidence than correlations, even if one is from social science and one is from genetics.

  5. Here’s my post on this very topic from late March:

    Cryptic Admixture, Mixed-Race Siblings, & Social Outcomes.

    My conclusion is that the necessary scientific will does not exist for researchers to organize the right samples and measure the correct variables, and that existing datasets will not have the precise variables needed to resolve the issue.

    My co-bloggers at Human Varieties have done a number of sibling pair analyses in the NLSY and Add Health datasets, investigating the relationship of skin color differences on IQ. Lighter skinned blacks perform better than darker skinner blacks on IQ tests, but this relationship only exists between families, not within families. Similarly you can look at the nexus between IQ, self-reported ancestry and interviewer-rated appearance. Once again, with this method, ancestry is what mediates the relationship between skin color and IQ. It’s not socially mediated, because the relationship is not driven by appearances.

    • toto

      “but this relationship only exists between families, not within families.”

      If there is any linkage between appearance and “IQ” loci, would this not point towards a mostly non-genetic explanation for the black-white difference?

      IIUC the recent occurrence of admixture, and the polygenic aspect of “IQ genes”, both increase the probability of linkage.

      • razibkhan

        If there is any linkage between appearance and “IQ” loci,

        very unlikely because of the different genetic architectures. appearance (skin color, nose shape) seem to be controlled by a much smaller number of genes than intelligence or height. if every appearance locus was flanked by IQ loci still would not be many IQ loci.

        i think the point here is that within family skin color is a good proxy for ancestry, and that family IQ variance is already constrained in comparison to the general population.

    • Emil Kirkegaard

      Arthur Jensen suggested this test for pleiotropy of skin color and IQ. Apparently no pleiotropy.

      Cite: Templer, Donald I., and Hiroko Arikawa. “Temperature, skin color, per capita income, and IQ: An international perspective.” Intelligence 34.2 (2006): 121-139.

      The commentary to this article by Jensen.

      • Richard Seiter

        The comments at the end of that article were an interesting read. I was especially intrigued by the author’s statement: “Major concessions were made to the extremely critical reviewer.”

      • razibkhan

        the genetic architecture for skin color is known (not too many loci, many of large effect). the genetic architecture of IQ is also known. many loci (many, many) of small effect. as i’ve told ppl a million times think about how that might work together in terms of pleiotropy. it can’t. there *could* be pleiotropy in IQ and height, because their genetic architecture matches.

        when i told rushton this privately in email he basically brushed it off and ignored me and kept talking about the possibility. he did this on some other clear genetic questions where the model he was promoting was obviously wrong. not the first time an academic has ignored disconfirming evidence, but still.

        • Emil Kirkegaard

          Yes, you are right about that. It could not be a large effect. Even if IQ and skin color had pleiotropy, it would be a very small effect on IQ (only 6 genes or whatever).

          In any case, someone did test whether skin color mediates IQ in siblings. They found that it didn’t. So pleiotropy also appears to fail the empirical test, along the the discrimination theory based on skin color.

  6. razibkhan

    With current technology could we not isolate the European fragments in the DNA of a US Black sample and learn something about the source(s) of the European DNA?

    well, yeah, in fact, you have plenty of AA american samples around. just pull out euro segments and see where they plot on the PCA, right? also, if they are mostly jeffersons then the effective pop is probably small.

  7. Ron Strong

    It might be quite difficult to do the kind of study you suggest, not for any technical reasons but for cultural and political reasons. Good random genetic samples from the target group (African Americans) would be required. This may not be easy to obtain.

    There are legal and “ethical” considerations that prevent sampled tissues or the data obtained from them (DNA sequences) from being used for any purpose other than those explicitly authorized by the donor. Subjects would need to be told that the study involved would help to determine whether there is a genetic basis to the IQ gap between blacks and whites.

    Blacks tend to be reluctant to participate in any kind of DNA testing. I would think that only a small minority of blacks would volunteer to provide their DNA for purposes of determining whether there are genetic bases to the IQ gap between blacks and whites.

    Self selection by the studied group presents obvious problems. There would be a high likelihood of systematic biases – those who volunteer are likely to be atypical of all African Americans. One certainly could not just advertise for submissions of specimens and academic achievement data. Given the hostility of many towards any research concerning genetic basis of IQ, falsified submissions aimed at corrupting the research would be a major issue.

    As an aside, do you have any idea how the BGI Gene-Trait Association Study of Intelligence handles the issue of intentionally false data submissions? I could certainly imagine John Horgan ( ) and many others like him submitting high IQ data along with the saliva of some bum off the street.

    • razibkhan

      Good random genetic samples from the target group (African Americans) would be required. This may not be easy to obtain.

      you can hundreds from the hapmap by downloading them right now.

      the challenges exist. but there’s going to be lots of data in the next 10 years leaking about.

      • Getting the genetic sample should be easy, it’s true.

        But how do you get in contact with the hundreds of sibling pairs you collect, and make them take IQ tests?

        • razibkhan

          if i was a wealthy person curious it might not be so hard. just pay people like universities pay college students.

    • Emil Kirkegaard

      One could pay them for the samples. Don’t mention that it will be used for only IQ testing. Tell them it will be used to look for all kinds of differences on a population level, and between population level. That is, after all, true. Then give them all kinds of tests, ask about disease history, interests, personality test (big five).

  8. > Obviously many people who write about this topic do not adopt a uniform prior. [my emphasis -Razib]

    Which raises the interesting question: what prior *should* we have? It’s not as if we are in a position of maximal ignorance where a uniform prior is remotely reasonable. We’ve observed many human traits, we’ve checked for lots of genetics or heritability measures, we do have base rate information on how likely a priori IQ being genetic ought to be. Do we have anything like “One Hundred Years of Social Psychology Quantitatively Described”,%202003.pdf Bond et al 2003?

    • Emil Kirkegaard

      Jensen’s default hypothesis?

      • That’s just an issue of dueling priors. What I’m asking for is an empirical prior, based on explicit sampling of scores or hundreds of studies (or ideally meta-analyses) and extracting their hereditability.

        (You might consider this as a sort of
        ’empirical Bayes’ approach to having priors for scientific matters: we pretend we have no prior and simply compile one from the field as a whole.)

  9. Race is not specific enough and can probably only rarely be used. Most human gene pools have not had enough isolation. Also ‘races’ are still evolving. There are always new mutation so when does a race split and become a new race? This needs to be addressed first before any “scientific” research can be done.

    • razibkhan

      this is a contentless comment. improve in the future or they won’t get posted.

  10. omarali50

    There was some discussion of this topic on (a liberal blog with a Western-Pakistani bias) when they posted Razib’s earlier post about biology and race and a post from Atlantic magazine ( Most comments were in the category of “I am shocked anyone would even think such thoughts in this day and age”, but a commentator asked this question and I am curious if anyone has a detailed answer (or a link to a detailed answer): How much doe this matter? Suppose we find that the average Japanese is a few IQ points smarter than the average Indian. What then? In terms of policy or individual life decisions, what will that change?

    This is not a rhetorical question. I am curious to know what people think..

    • if we give an answer we’re then called racist. it’s a good question but i typically decline to answer those because sometimes the answer isn’t a nice one.

    • If that is true, some arguments in favor of affirmative action (that having a workforce which does not accurately reflect the makeup of the population of the nation as a whole shows evidence of prejudice) are false. Note that there are still other arguments in favor of affirmative action even if all groups are not equally capable. For example, one could argue for it in terms of reparations (although then it would only apply to groups who historically suffered, like blacks and native Americans), or that diversity is good even if it trades off some competence (which I’d argue is the case for at least some fields, like public policy and government in general).

      I would argue that even in a hypothetical world where we have knowledge that different groups tend to have different intelligent levels, we shouldn’t stereotype about abilities – that we should judge each person as an individual. I have seen HBD proponents argue in favor of racial stereotyping, on the grounds that it’s more likely to be accurate than not, which is one of the reasons why that “community” makes me edgy, and I believe a lot of people involved are just looking for reasons to justify their priors (as, admittedly, I am as well).

      • Emil Kirkegaard

        Racial stereotyping would be useful in cases where individual testing is not available. I dislike group discrimination and like individual discrimination. This is important for e.g. immigration policy.

        • I’m not convinced at all that IQ scores can be compared across countries of different development levels, so the utility in terms of immigration is somewhat limited.

          FWIW,if it became proven most of the black-white IQ gap was hereditary, I actually think that promoting interracial marriages to the highest degree possible (although obviously not compulsory) would be of paramount importance. It wouldn’t affect averages across the U.S., after all, but could within a generation or two at least dampen the social salience of race in our culture – and I think heading off a return to gross stereotyping is important in order to maintain cultural cohesion.

          • Emil Kirkegaard

            It would boost scores slightly due to heterosis (hybrid superiority).

            And yes, your proposal might lead to less social problems. For a very homogenous country, like mine sort of, but decreasing fast (Denmark), this is a bad idea. Better to have strict immigration laws.

          • Gregory Cochran

            Some time two populations ‘nick’ and show heterosis: sometimes they don’t.

          • Emil Kirkegaard

            Any reason why? I remember reading about a heterosis effect in mixed US white+hawaiians.

          • razibkhan

            i suspect that in most ‘heterosis’ cases it is just masking of mildly recessive genetic load. this means that if F1’s mix you should lose some of the benefit in F2 because it segregates back out….

          • Gregory Cochran

            It would also decrease the number of people with very high IQs.

          • Eh. I think that it’s worth trading off a few very high-level geniuses in exchange for having a mixed-race, largely homogeneous society where there is no danger of a racial caste system evolving.

          • razibkhan

            you need to do more than admix. caste systems will emerge through assortative mating if physical appearance has a social value/valence. it is not too hard theoretically to extract ‘whites’ and ‘blacks’ out of a 100% mulatto population

          • The nonwhite population in the U.S. is low enough (particularly when you consider the white admixture in blacks and Latinos) that someone who was a completely average American would be more than 80% white, and only around 10% black, with the rest various Asian/Native American admixture. That’s barely at the point of detectable ancestry.

            Of course, in the real world, there would be a color gradient due to assortive mating, as you suggest. Still, I think having everyone a little more mixed has positive social value in somewhat easing differences, given other “solutions” are just horrifying.

          • razibkhan

            i see. you are then think of an argentina model. about ~15-20% of the genome of the nation seems non-european, but it self identifies as a white settler nation since it’s phenotypically cryptic.

          • Gregory Cochran

            We can be a fountain of progress like Brazil.

          • On the other hand, India throws off its share of geniuses, considering the overall IQ is fairly low and people there are quite admixed.

            Regardless, I see myself as an American first and foremost, not a white person, or even a smart person. I just don’t see the worth of a society being comprised by how many “great people” it creates, but the aggregate well-being of the majority of the population.

            But then again, as I said, I’m a socialist in terms of my economic outlook, so it would make sense that leveling, all things considered, is more appealing to me than the alternatives.

          • Gregory Cochran

            “admixed” in the sense that major castes have been reproductively isolated – and subject to different selection pressures – for about three thousand years.

    • razibkhan

      remember that a small average difference can have huge tail effects.

    • Anthony_A

      Charles Murray, in The Bell Curve, said that the persistent difference in IQ between blacks and whites, plus the history, justified *some* affirmative action – he suggested a bump equivalent to half the difference.

      Once the fact that racial differences in IQ are genetic is more widely accepted, we can stop harassing teachers for being racist and perpetuating the test score gap. We can also recognize that our (American) teachers are actually doing a pretty good job compared to their predecessors, once we account for the change in the student population.

      We might even get some reversal of white flight, if people recognize that differences in school test scores are largely artifacts of different demographics in the schools, and that any school which is reasonably orderly and safe is likely to be as effective at teaching any particular kid as any other. In the town where I live, test scores vary widely across schools. Test scores for white and black kids vary across schools almost not at all, and for Hispanics and Asians, not that much. Thus I don’t worry about the academic quality of my daughter’s school – it does just as well with white kids as do the schools with more whites and Asians and higher test scores.

      • Gregory Cochran

        I thought I was the only person alive who believed that enough to put his children in the local mostly-minority public high schools.

        There is another !

        • Anthony_A

          Well, not “mostly-minority”, even if you include Asians, though probably majority-minority if you include Asians. And there are non-academic issues my wife and I worry about. If we lived in Oakland, those issues would be enough for us to deal with private school.

    • Emil Kirkegaard

      It means we should not use affirmative action for the reason that certain minorities do badly because they are discriminated against.

  11. omarali50

    btw, admixture mapping has been done for obesity and related traits and did find that higher European ancestry correlated with lower obesity and diabetes rates in African-Americans (and in an earlier study using fewer markers, in Pima Indians)

    Also works for uterine lyomyomas

    • Emil Kirkegaard

      Lower obesity, also a known correlate of g factor. Another confirmation for hereditarian position.

  12. Gregory Cochran

    Differences in average brain volume between groups might affect your priors, except for the fact that our technology just hasn’t advanced to the point of being able to accurately measure things the size of a bowling ball.

    Although we do ok on protons.

    • If brain size really made a dramatic difference in IQ, then why don’t women have significantly lower IQs than men do?

      Further, I’m pretty sure that some Arctic groups have the highest EQ of all (Eskimo, various Siberians), but tend to test around 90 on IQ tests.

      • Emil Kirkegaard

        Yes, this is odd. I have some proposals. Research is definitely ongoing in the sex question. Have you been following it? It is very interesting. IQ cannot be relied upon, need to look at g-scores.

      • Gregory Cochran

        All hunter-gatherers have IQs below 100 – but Eskimos have the highest. It is as if brain size if one of several factors affecting intelligence.

  13. razibkhan

    and I don’t see any easy ways around the natural cofounds between environmental and genetic effects, particularly given that heritability of IQ is lower at lower SES than at higher SES.

    look, how about comparing high SES black siblings where the ‘whiter looking’ sibling is less white genetically? all the institutional racism models i’ve seen would indicate that these siblings would perform better or the same on psychometric tests than their ‘blacker looking’ more white sibling. there are plenty of these people, so it’s a matter of finding them. or, you could just go to cape verde and test lots of families which match the pattern, since the vast majority of the population are a recent admixture.

    you prove that intragroup variation is much more important than between group variation which supports the status quo on how we treat group differences in IQ in the U.S.

    the null model is that NO INNATE differences exist. so the target would differ, if only somewhat. people have been known to expend inordinate resources on-the-margin, so it might be important to know when to ‘stop’.

    • andrew oh-willeke

      * The trouble with siblings is birth order confounds and when one is talking about “whiter looking” how to quantify a continuous but ill defined variable. A case like Cape Verde is subject to selection issues related to the settlement and socioeconomic history of the jurisdiction. Getting a non-biased sample is so hard that unless you have an intrinsic way of adjusting for it, systemic uncertainty is significant, undetermined and not easy to rule out since any significant effect could be due to some bias in a factor not measured by the researchers. The other issue is that “non-shared environment” in heredity estimates is more of a junk “what’s left over” category than one that is meaningfully theorized and understood in any comprehensive way.
      * The null model is probably more precisely that no innate differences in predicted IQ base on genotype alone that can’t attributed solely to statistical sample error exist, on average, between groups of different races that are sampled in an IQ neutral manner.
      * I don’t hear anyone arguing strongly the genotypes that lead to average predicted IQ based on genotype have no ancestry informative differences – merely that the differences are a wash. Also, realistically, “exact” equality isn’t really what people care about. It is virtually impossible to get “exact” equality in any model where multilocal genotype leads to a predicted IQ in populations of millions of people, even though it can be arbitrarily small. For these purposes arbitrarily small (e.g. smaller than any existing testing instrument can distinguish in an individual) is surely good enough and those differences amount to no more than random noise.

      • razibkhan

        A case like Cape Verde is subject to selection issues related to the settlement and socioeconomic history of the jurisdiction. Getting a non-biased sample is so hard that unless you have an intrinsic way of adjusting for it, systemic uncertainty is significant

        all samples are somewhat biased. your objection strikes me basically as “sample size too small!” complaint. it’s theoretically legitimate, but you’ll always complain the N wasn’t large enough and the sample wasn’t representative enough. you might be happy though if N == whole human population 🙂

  14. botti

    ***though the actual authors reject that interpretation***

    @ facefault,

    Note Sandra Scarr’s subsequent comments in a 1998 article.* They demonstrate why this area is difficult to research and the pressure on researchers to find environmentalist explanations.

    “My colleagues and I reported the data accurately and as fully as possible, and then tried to make the results palatable to environmentally committed colleagues. In retrospect, this was a mistake. The results of the transracial adoption study can be used to support either a genetic difference hypothesis or an environmental difference one (because the children have visible African ancestry). We should have been agnostic on the conclusions…”

    * Scarr, S. (1998). On Arthur Jensen’s integrity. Intelligence, 26 (3), 227-232.

  15. razibkhan

    I ASKED YOU SPECIFICALLY ABOUT THE QUESTION IN REGARDS TO A SIBLING DESIGN. answer that question, not in the generality. how does gene-environment interaction work differentially with siblings in the case where cryptic ancestry varies.

    i’ve read falconer & mackay, please don’t lecture me on gene-environment interaction. don’t assume your interlocutors are so stupid that basic variables aren’t part of their background assumptions.

  16. Emil Kirkegaard

    It DOES matter. Because it means we should not have affirmative action and other disgraceful discrimination based on the idea that social differences cause the differences we see. Clearly, positive discrimination for blacks in the US have not made them as smart as whites. Worse, it means that the black average at universities will be lower than the white/asian, just like in regular society.

    • I don’t understand the argument affirmative action is disgraceful. If it doesn’t work as intended, it may be inefficient. Certainly if our goal is to maximize productivity, there are better ways to organize the workforce (even the most “nurture”-oriented people regarding IQ admit it is closely associated with job performance. But I don’t see how it makes sense to add a moral judgement to it.

      But I didn’t end up with a middle-class salary because of hard work. I ended up as such because I inherited genes for above-average intelligence from both my parents, which resulted in me doing well in college and graduate school, which then resulted in me getting a professional job. Someone else may have gotten into the same position due to affirmative action. Maybe they are less competent (or maybe they are more so in a different way – I have pretty low conscientiousness) – but that doesn’t imply they worked any less hard. It doesn’t imply anything regarding who deserves what.

      That’s why I don’t see why people (often in the HBD community) get so worked up about affirmative action. I suppose I can see if you come from a narrow procedural idea of justice. But understanding how the human brain (and genetics) works makes it clear there are no “just rewards.” In a certain sense, pure meritocracy is no different from aristocracy, as it gives people high positions due to “birthright.” Obviously the main difference is unlike aristocracy, meritocracy has much higher results in regards to productivity, but other questions besides productivity (such as general social cohesion) can play a role as well when deciding how to slot people.

      • razibkhan

        I ended up as such because I inherited genes for above-average intelligence from both my parents, which resulted in me doing well in college and graduate school, which then resulted in me getting a professional job

        also, i think we do underestimate luck/contingency in getting to where we are. huge variance in $ among high IQ people, and not just because of different professional paths (e.g., academia vs. hedge funds). see taleb’s critique of ‘millionaire next door’

      • razibkhan

        I don’t understand the argument affirmative action is disgraceful.

        i think it is offensive when high socioeconomic status people of color tend to gobble up most of the benefits of positive discrimination. everyone has plenty of examples. a friend of mine is half latino and half scandinavian. she takes after her scandinavian side in looks, does not know spanish, and her parents own a vacation home in a spendy locale (they run a successful business). but he obtained a nice minority fellowship for herself. i am happy for her because she is my friend, but it seems rather disgraceful, because i doubt that the spirit of these programs was originally to benefit people like my friend. but of course it is highly unlikely that the mexican american child of laborers would be in a position to obtain a graduate fellowship in the first place, so that’s what happens.

        • I actually do have issues with Affirmative Action, but not the same issues as most people on this forum, but in some ways similar to what you note here.

          As I see it (coming from my own political stance, which is broadly socialist), the central problem regarding the African-American condition is poverty. Affirmative Action was a dubious solution to this, because even if it worked as intended, it only helped a few “lucky duckies” who chanced their way into the system – not the large poor black underclass. Indeed, on a fundamental level, I think Affirmative Action is no different than school vouchers, despite proponents generally being on opposite sides of the spectrum. Both try to solve systematic problems by giving a small subset a leg up. But helping the “talented 10%” doesn’t help the other 90%.

          In some ways, I think the bigger social effect of Affirmative Action within the U.S. regarding race has been on white people. For example, while only a few blacks in an office don’t make a major difference, it probably over the longer run changes the opinions of many of their white co-workers, who otherwise would never socialize with someone black. I think that the lessening of stereotypes has been an unmitigated good, so this aspect of Affirmative Action has not been unsuccessful.

          Still, I’d be all for scrapping Affirmative Action in exchange for say having a real national policy for creating good-paying jobs in the inner city – which would help the status of African-Americans over the longer haul far, far more.

  17. Emil Kirkegaard

    At least one of the authors does not reject it. She explicitly tried to make the data palatable to the environmentalists at the time, a move she regrets.

    “An interesting parallel to this work is our longitudinal study of interracial adoptees. At the average of 7 years, the African-American adopted children scored 106. I on IQ tests. By the average age of 18 however, their IQ scores had declined to 96.8. Children with one White and one Black parent scored, on average. 109.0 at age 7 and 98.5 at age 18; children with two Black parents (and later adoptive placements) scored 96.8 at age 7 and 89.4 at age 18. The test performance of the Black/Black adoptees was not different from that of ordi- nary Black children reared by their own families in the same area of the country. My col- leagues and I reported the data accurately and as fully as possible, and then tried to make the results palatable to environmentally committed colleagues. In retrospect, this was a mistake. The results of the transracial adoption study can be used to support either a genetic difference hypothesis or an environmental difference one (because the children have visible African ancestry). We should have been agnostic on the conclusions; Art would have been.”

    See: Scarr, Sandra. “On Arthur Jensen’s integrity.” Intelligence 26.3 (1998): 227-232. The entire issue is devoted to Jensen.

  18. razibkhan

    with the environmental component ubiquitous in the US (take your pick), but e.g. not in Africa.

    this is why it would be useful to check across a variety of world populations. cape coloreds, cape verde, latin america. the project could scale pretty easily.

    • genobollocks

      Ok, so we agree on GxE. Though people would always find ways out of correlational evidence, e.g. by arguing that those who have more Black ancestry respond more aggressively/defiantly to discrimination, which again would be shared among all those mixed populations you mentioned. But that would seem pretty ad-hoc.

      You’re now bringing up a different criticism of ST research, namely that it generalises to policy interventions.

      Before, you claimed that it was a spurious result based on publication bias (as did Emil below). This criticism and the ones in the Wikipedia section are refuted by the meta-analysis I linked. Of course it’s not uncontroversial, but check out the evidence yourself, you might be convinced. You brought up priming repeatedly, but priming need not be the mechanism for ST, though that has been suggested, it could well be conscious rumination. I do not agree that priming effects are more traditionally accepted in psychology.

      The experimental claims of an aggregated effect size of .43 don’t seem so outrageous to me and I’m willing to bet 100$ on a pre-registered, adequately-powered RCT on ST conducted by able people finding an effect size > .10 in the same direction.

      However, I’m not going to bet on policy effects. What I invoked the stereotype threat example for, was to argue that there is evidence that should shift one’s prior towards some variance being environmentally accounted for, because it is experimental evidence in a realistic, relevant setting (results in tests taken do determine later life outcomes).

      That does not mean that a writing intervention supposedly boosting self-esteem has to work or that the whole gap should disappear when the ST is somehow lessened a little.
      But it goes to show a plausible, environmental explanation how test results can be lower in objective tests for a certain, stigmatised group, that is actually experimentally demonstrated. So, it beats a lot of the correlational evidence cited by those whose priors peak at more-genetic variance.
      If those experiments can show a small effect, the true effect of real, prolonged discrimination may be large. I think that is not a big leap. But you can’t experimentally expose people to real, prolonged discrimination.

      So, is that bet on? We would have to wait for someone to pre-register a stereotype threat study on eg. openscienceframework, but with the current mood in psychological science, that will happen sooner or later. Betting on a non pre-registered study wouldn’t be fair to you.

      • razibkhan

        by arguing that those who have more Black ancestry respond more aggressively/defiantly to discrimination

        this is why i am interested in focusing on dyads where the sibling appearance and ancestry exhibit discordance. this would not be difficult to find, especially in the 40-60% ancestry interval.

        re: bet. i’ll have to read the most recent literature first. i also have a lab presentation end of week, so can you ping me on twitter in a week to push me to do this in case i haven’t gotten to this weekend. $100 is a manageable amount though, so if i still am unconvinced i’ll be game.

        and to be clear: i assume that at least some of the variance in group performance is environmental, particularly due to various forms of gene-environment *correlation*. the issue is on prior grounds do you believe that ALL of the gap on psychometric tests is due to environmental factors? i’m moderately skeptical of that assertion.

        • genobollocks

          Re: the dyad case. Yes, the most obvious thing would be appearance, which you could rule out via discordance with actual ancestry. But what I suggested (actual ancestry-aggressive correlation) would work (because it is an ad-hoc idea to highlight a limitation).

          I could certainly be convinced of a genetic component with good evidence. So my prior on that isn’t zero or extremely low. There’s room for both explanations.

          But I haven’t seen good evidence. I can’t say that Lynn’s or Rushton’s studies were good tests of their hypotheses, but the same can be said for proponents of the other side (see recent Neuron article purporting to not find g). But that’s why I expect your proposed study design not to be carried out in its optimal form. Most actors are too ideologically entrenched to disconfirm their own preconceived notions.

          • razibkhan

            I could certainly be convinced of a genetic component with good evidence. So my prior on that isn’t zero or extremely low.

            re: quantitative traits, my *evolutionary prior* is that there is likely to be some group differences on many complex traits if you could magically control environment. this seems likely for height. the differences are modest across most populations (e.g., perhaps 2 inches between europeans and east asians, greater if you subsample europeans to only northern europeans; smaller if you compare north/south europeans

            Most actors are too ideologically entrenched to disconfirm their own preconceived notions.

            the cost of genotyping (not sequencing necessarily) is going to be trivial. there will be national databases. the military in the USA does the ASVAB test for all recruits, and it may in the near future also genotype them (i had a friend who died of an adverse reaction to the anthrax vaccine after basic training, so there are reasons to understand genetic risks). so they could do it. of course, there’s the issue of will, but i am not optimistic that data leaks will not occur.

  19. JL

    Stereotype threat manifests itself as a violation of measurement invariance. However, there are generally no violations of measurement invariance in studies of black and white American IQ differences (for example, see this study and the references therein). This indicates that stereotype threat is a laboratory curiosity with no real-world significance. The real-life black-white gap is 1 SD, but stereotype threat can make it larger in some settings.

    • The black-white SAT gap is larger than the black-white IQ gap.

      I mean, I’m math challenged, but under the new SAT system, the average white student scored 1579 in 2011-2012. Presuming a perfect correlation with IQ (e.g., blacks score 85% of white scores), you’d expect the average black score to be 1342. But instead the average score is 1272. Thus around 70 SAT points are lost in addition to what would be expected from tested IQ. Thus there’s clearly some underscoring for some reasons (socialization, stereotype threat, whatever) which even the most doctrinaire HBD supporters would presumably agree we should be able to deal with.

    • genobollocks

      I don’t see how that follows. Wicherts, Colan, & Hessen don’t seem to think so either. Do you mean to say this invalidates the meta-analysis I linked and my account (to Razib) of why this should shift your prior? I understand the authors as giving one theoretical explanation of stereotype threat affecting measurement by especially affecting difficult items? As far as I understand it this statistical property needs to be evaluated in context, it doesn’t just mean “these results are bollocks”. Do you want to elaborate what you mean?

  20. genobollocks

    Sorry, this is just dishonest argumentation. I showed you a meta-analysis that includes 43% unpublished studies and assesses problems related to publication bias, comes to the conclusion that ST exists and you link to some speculation on Steve Sailer’s blog? John List may be an a bit more respected academic than Steve Sailer, but he still isn’t above the literature with his armchair speculation and economics exceptionalism.

    I know what measurement invariance is and means, apparently you don’t. Or maybe you still misunderstand my argument. I’m not saying that ST during IQ testing explains the B/W test score gap in its entirety and that in reality blacks should have much higher IQs.
    What you say does not invalidate the notion that ST is a plausible mechanism that can link discrimination to lower test scores. Test scores work as gatekeepers at many levels of society. For example they can affect how much education you get. Education is bidirectionally linked to IQ. Some of that and other effects (discrimination in hiring, thus income etc.) may explain some of the difference in mean IQs. Maybe you are equating latent variables with “the true underlying real potential intelligence” and thus mean to argue that the gap is in real underlying potential and could not be affected by something that does actually lower real underlying potential. If that were so, you’d be confused about what latent variables are and aren’t, I’d recommend reading up on Denny Borsboom’s work.

    I also take issue with this statement: “ST is a form of test bias that can be detected statistically, without any need for experimental designs”. Though I didn’t argue for this explanation for the IQ score difference, that statement is a plainly wrong oversimplification. Experimentation is used to isolate effects. If ST were the only factor that could possibly affect measurement invariance, this would be true. It is not. There may be for example SAT prep course trainers that have noticed that their black protégès fare better if they tell them to tackle difficult items first or last or whatever. It might counteract, diffuse or strengthen measurement invariance caused by ST. ST may still decrement test scores, though differently, with such a moderator in place.

    Additionally I’d like to say that judging by the sources you chose, I’m not surprised that you say that every study you’ve seen validates your preconceived notions. If you’re interested in the truth, you should maybe change the way you look for evidence.

    • JL

      You are correct that a violation of MI can be caused by other forms of test bias than ST, but the point is that if ST was a significant factor in the B-W gap, we wouldn’t find MI in analyses of black and white IQs. But we do find it.

      You appear to naively think that if you find some effect in artificial experimental settings, it means that the same effect must be operative in real-life situations. But that’s of course hogwash.

      There’s no evidence that black ability is underpredicted by intelligence tests. If anything, it’s overpredicted. Of course you could argue that there’s a pervasive ST effect everywhere in American society causing blacks to underperform to the same degree in all tests, all schools, and all jobs, generation after generation. But that would be just silly and in fact not what the ST paradigm predicts. ST should be highly situation-specific and fickle, not some constant that is present everywhere.

      BTW, because of affirmative action and the disparate impact doctrine, test scores work as gatekeepers for blacks to a lesser degree than for whites and others.

      Additionally I’d like to say that judging by the sources you chose, I’m not surprised that you say that every study you’ve seen validates your preconceived notions.

      Who do you mean? Jelte Wicherts? Conor Dolan? Gitta Lubke? All are well-respected psychometricians who have written a lot about latent variables (in fact, all have cowritten papers with Denny Borsboom). Can you present any specific criticisms of their research? Wicherts, incidentally, has a meta-analysis under review where he reportedly argues that ST literature suffers from publication bias and that ST is pretty much non-existent in African Americans in most settings.

      If you want to challenge my arguments, show me that ST is a significant factor affecting the cognitive performance of African Americans in real-life settings. If you know studies showing measurement non-invariance between whites and blacks, please name them. But I suspect that you’re full of it.

      • genobollocks

        I was not referring to the psychometricians, but to Steve Sailer and should have made that clear. I thought you were trolling at this point, though now I don’t think so anymore.

        I think I have challenged your arguments sufficiently and will not let you set the topic, so that I’d have to defend a position that I never held. I will reiterate once:

        ST experiments demonstrate a plausible mechanism by which discrimination can lower test scores. You could call these scores “cognitive performance” – what violation of measurement invariance is about is that one comes to different scores differently (for example affecting the parameters of a difficult subtest), the performance is still lower as measured by simple sum scores.
        How about this quote: “Clearly, the relationship between test score and latent ability now depends on group membership, and the requirements for measurement invariance no longer hold. Therefore, stereotype threat effects are by definition a source of measurement bias.”
        This should tell you that the authors you cite do not think violation of MI negates the validity of ST. I don’t know what you’re trying to argue by saying that these authors wrote a lot on latent variables, together with Borsboom. Yes. Does this mean that latent/g=real (the misconception I ascribe to you due to which I thought you didn’t know this literature)? No.
        Because simple sum test scores function as gatekeepers, your caveat notwithstanding, that may mean less access to education, jobs etc. These are plausible mechanisms for a depression of real intelligence (see e.g. IQ decrements during summer vacation).

        You also misunderstand me when you say that I believe “the same effect must be operative in real-life situations”. I think that experiments can show us whether mechanisms are plausible. ST may be fickle, but it’s only a subtle, ethically defensible, intervention. The effects of prolonged, life-long discrimination could be stronger and I’m not just speculating on thin ice here, simply read the intros of the articles you linked.

        I read Wicherts’ introduction of the problem of generalizability to high-stakes settings, so as if he was undecided whether the reason they do not find it there, is that it is hard to increase stereotype threat further for one randomised group when both groups are already stereotype-threatened through high-stakes.
        You seem to hold the false conviction that different kinds of bias sources can only add up to more (measured) bias. This is false, two sources can also cancel each other out.
        That said, I understood Wicherts et al. as saying that MGCFA could be used to investigate ST in high-stakes settings (where it may be so that both groups were under some ST), not that MI in B/W IQ differences meant that stereotypes had nothing to do with the difference.

        You mention a meta-analysis by Wicherts under review. If you have it, you can send me (and Razib, since we’re betting, for fairness’ sake) a copy to this moniker at hushmail, but otherwise we can hardly discuss it.

        • JL

          This should tell you that the authors you cite do not think violation of MI negates the validity of ST.

          You still don’t get it. The point is that if there’s MI, there’s no ST, and vice versa. And MI is found in all non-experimental studies of black and white IQ that I know of.

          I don’t know what you’re trying to argue by saying that these authors wrote a lot on latent variables,

          I understood that you were questioning their authority, so I pointed out that they are in fact well-known experts on just these topics.

          I don’t have the text of Wicherts’s meta-analysis. He and a colleague presented their results at a conference a couple of years ago. I learned about them from (at this point, make sure you have your smelling salts at hand) none other than Steve Sailer, one of whose readers was present at the conference. Here’s the abstract by Wicherts & de Haan:

          Numerous laboratory experiments have been conducted to show that African Americans’ cognitive test performance suffers under stereotype threat, i.e., the fear of confirming negative stereotypes concerning one’s group. A meta-analysis of 55 published and unpublished cognitive tests due to stereotype threat, this effect does not generalize to non-adapted standardized tests, high-stakes settings, and studies of this effect shows clear signs of publication bias. The effect varies widely across studies, and is generally small. Although elite university undergraduates may underperform on less academically gifted test-takers. Stereotype threat cannot explain the difference in mean cognitive test performance between African Americans and European Americans.

          • genobollocks

            > The point is that if there’s MI, there’s no ST, and vice versa.

            I get your point. It’s wrong. It’s also largely irrelevant to my line of argumentation.

            Thanks for clarifying your source. Let me summarise.

            My source is published, deals with publication bias by including both unpublished and published studies, finds a mean effect (which is small but sufficient to support my line of argumentation) and identifies moderators (some overlap with your abstract, none of which are problematic to my line of argumentation).

            Your source is unpublished and all you have is a conference abstract and hearsay to go on. In published work the first author does not agree with your claims and paints a more nuanced picture.

            I’ve contacted Wicherts to see what has come of it. In my experience, 4 years is a long time from conference presentation to publication.

            Probably it was suppressed by the PC conspiracy, right? Oh, but Wicherts is an open access pioneer. He couldn’t have possibly self-published or something? This couldn’t possibly mean that there were problems with the work?

          • JL

            Your argument is that in some unspecified circumstances the black-white gap is larger than usual due to ST. I doubt that this is the case in any societally important situations, but in any case there’s no evidence that the typical 1 SD gap, shown in hundreds of studies and data sets, would be any smaller even if there was no ST anywhere.

            Let me know what Wicherts has to say. I’d have expected that his study would have been published by now, too.

  21. szopen

    I am just reading Jensen’s “g-factor” and even thoug hI have not finished, I am already surprised how a lot of commenters in the thread under Coates’ interview have no idea what they talk about.


Discover's Newsletter

Sign up to get the latest science news delivered weekly right to your inbox!

Gene Expression

This blog is about evolution, genetics, genomics and their interstices. Please beware that comments are aggressively moderated. Uncivil or churlish comments will likely get you banned immediately, so make any contribution count!

About Razib Khan

I have degrees in biology and biochemistry, a passion for genetics, history, and philosophy, and shrimp is my favorite food. In relation to nationality I'm a American Northwesterner, in politics I'm a reactionary, and as for religion I have none (I'm an atheist). If you want to know more, see the links at


See More


RSS Razib’s Pinboard

Edifying books

Collapse bottom bar