Sunday 20 April 2014

Question time #LCI14

 

Would you like to ask some researchers questions about intelligence?

The sorts of topics could be: Spearman's Hypothesis examined for primate cognitive comparisons; Why don’t Northeast Asians win Nobel prizes?; The Roma: a Balkan underclass; Science and its discontents; The evolution of racial differences in intelligence, in psychopathic personality and sporting abilities; Sex differences in intelligence; Polygenic selection and human evolution; the intelligence of the Victorians; Understanding heritability estimates; the General Factor of Personality, Dysgenic trends in simple reaction times in Scotland and Sweden; Immigrant attainments in Denmark; cognitive ability in Mexico; g in dogs.

Leave your question as a comment on this blog, and I will try to get an individual researcher to reply to you, either directly or through me. If you favour brevity, then tweet me your question.

27 comments:

  1. No original questions here but I'd love to hear your thoughts on...

    > Why don’t Northeast Asians win Nobel prizes?;

    > The evolution of racial differences in intelligence, in psychopathic personality and sporting abilities

    ReplyDelete
  2. Is high IQ a mating or a survival advantage?

    ReplyDelete
    Replies
    1. Both. You solve the problem of survival, and she notices that her future looks better in your company. You get the life, and the girl. From her point of view, she gets protection and her children survive. Happiness all round.

      Delete
    2. Thanks, James, for the reply. I understand then it is first a survival advantage, and afterwards a girl notices and it becomes a mating advantage.

      Delete
  3. oh boy! over what time span (& at what rate) did racial group mean differences in intelligence evolve? what can we expect in the future? is it always a trade-off? (this group is higher on g, but lower in athleticism/speed, etc.)

    do dysgenic trends vary by group? do Flynn effect rates vary by group? (higher-g folks learn matrix rules more quickly & thoroughly, yes?)

    do dysgenic/eugenic trends vary by SES?

    are we creating more variance/larger standard deviations over time through assortative mating?

    what can we do about "IQ inequality?" (just kidding on that last one).

    ReplyDelete
    Replies
    1. To your first question, it would seem that clear advantages should be evident after about 16 generations. Try Westhunter on The Breeder's Equation.

      Delete
    2. thanks! here's H. Harpending on the matter, from his comment at http://scienceblogs.com/gnxp/2007/08/13/how-fast-do-gene-frequencies-c/
      (NB: substitute "IQ" for "time preference")

      "...For a ballpark (estimate), assume time preference has an additive heritability of 25%. Assume that everyone with time preference more that 1 sd above the mean... has double the fitness of everyone else. About 16% of the population then has twice the number of offspring as everyone else on average.

      After a generation of reproduction the new mean time preference will be increased by (0.2 * .25) or 5% of a standard deviation. In 20 generations, 500 years, time preference should go up by a full standard deviation."

      Delete
    3. That's a lot of questions. Evolution of race average IQ differences must have occurred in the last 60,000 years or so, because before that time our ancestors were still confined to some place or another in Africa, with a population size (judging by present-day genetic diversity) that was not very large. Later Neanderthal and Denisovan admixture was too small to make much of a difference. But we don't know whether the important factor was a harsh Ice Age climate, which required complex cultural adaptations and therefore high intelligence, or whether most selection took place after the neolithic revolution when population size (and therefore the supply of new mutations) was greater and demands on foresight in seasonal climates were greater. Either way, any evolution of racial intelligence differences must have been quite recent.

      Flynn effects do differ by groups. In the United States, for example, they were stronger for blacks than whites, at least for cohorts born between about 1955 and 1970. Flynn effects differ by country as well. Today, Flynn effects have ended in Britain and the Scandinavian countries, but there are Flynn effects in many (almost certainly most) developing countries. Dysgenics is harder to track because its real-world effects are masked by Flynn effects, which are of far greater magnitude. Generally, dysgenics seems to be smaller in more egalitarian societies (Scandinavia) than in highly stratified ones (Latin America). Also, differential fertility by education, at least, tends to be strongest during the demographic transition and somewhat smaller in societies with sub-replacement fertility.

      I cannot even guess whether dysgenics is greater in people of high or low SES.

      Assortative mating has always tended to maintain greater dispersion of ability in the population, but contra Herrnstein & Murray there is no evidence that homogamy by education or intelligence has been increasing over the last decades.

      What we can do to reduce IQ inequality is obvious. First, make sure that everyone has the same opportunity to develop his intelligence. This has by and large been done already through compulsory public schooling in advanced societies, but not in many developing countries. So, there is little more we can achieve through traditional measures in advanced societies. Therefore we need to equalize genes as well as environments. This has not yet been attempted and can therefore produce large effects. Ideally, we should outlaw sexual reproduction and clone every child from the same cell line, taking care not to introduce new mutations in the process. Second best measure: Find rare genetic variants with large effects on IQ. These will in almost all cases be IQ-reducing mutations that are rare because they are subject to mutation-selection balance, except perhaps in modern societies with dysgenic fertility. We need to get rid of these mutations! This will greatly reduce individual differences in IQ, but with little or no effect on average race differences, which appear to be caused by common polymorphisms with small effects. People have to realize that they don't need the yucky old-fashioned method anymore to make their children. We have test tubes! In-vitro is the most effective way to introduce quality control into human reproduction because every embryo's genome can be sequenced, and you simply use only the embryos with the least amount of genetic garbage. This is not only for social engineering. There are compelling humanitarian reasons for this procedure, despite diverse "ethical" objections.

      Delete
    4. Thank you for those excellent, insightful & enjoyable to read ("we have test tubes!":) comments - impressive answers, all.

      & definitely, if environments are made "equal," the only reason left for people to vary is genetics:)

      now, when did evolution of american-football differences occur? (kidding:)

      Delete
    5. "Therefore we need to equalize genes as well as environments."

      Variability in this trait doesn't have a species- level adaptive value?

      On the broader topic, I was imagining -- along the lines of Julian Savulescu --- increased, not decreased, differentiation.

      Delete
  4. Everyone talks about regression to the mean. What about the inverse? Suppose there are a thousand of IQ 140 for every one of IQ 165. Is it the case that sliding the IQ of the one thousand persons up and down and similarly for the one at 165, is there a point at which the person at the higher IQ is at a fifty-percent probability of being just as likely to be the offspring of two parents in the pool of the many as an offspring of two parents of the few? (In my jottings by the way, I refer to this opposition to the regression to the mean as 'progression to the extreme'.) I have a vague recollection of reading a comment by an intelligence researcher many decades ago that he had been puzzled for a long time that about half of the eminent minds he had studied, Nobel Prize winners and the like, had fairly ordinary parents. His moment of epiphany was when he realized that for every pair of extremely bright parents there are thousands or tens of thousands who are not. Genes do shuffle around quite a bit even in a single generation after all, and complex traits tend to draw on many genes as part of their expression.

    ReplyDelete
    Replies
    1. Very good question. I did a quick version of this calculation on the specific but relevant topic of university entrance by social class. That will give you the first step in the answer. I did not look past 2 sigma, but I could extend it to 3 sigma. At 2 sigma the differences are such that professional classes provide 5 times more bright people than manual labouring classes. I do not think that the differences between 2 and 3 sigma are a thousand fold. These two posts provide the beginnings of an answer, but I will have to think about it further.
      http://drjamesthompson.blogspot.co.uk/2012/11/social-class-and-university-entrance_28.html
      http://drjamesthompson.blogspot.co.uk/2013/12/the-7-tribes-of-intellect.html

      Delete
  5. First off, I should note that clicking the "preview" button simply makes your written comment disappear. So I'll try to recreate yesterday's.

    This may be somewhat basic.

    There seems to be widespread agreement that g is more interesting than IQ, the latter being merely an imperfect indicator of the former. For example, it is sometimes argued that the Flynn effect is not all that interesting, because it is not g-loaded.

    If you give a sample of people IQ tests, it should not be all that hard to calculate g values for individual participants. First, you give the test, then, using factor analysis, you determine the g loading ("load"?) of individual items, then you sum, for each subject, the number of correctly answered items, weighting them by their g loading. Alternatively, you could take the weights from a different sample, such as a study of the general population.

    Yet when I see studies which use intelligence as a predictor (say, of health outcomes in research by Gottfredson and others), the variable used is always IQ, not g.

    Why?

    ReplyDelete
    Replies
    1. Sorry about the preview button. I have never used it.
      You are right that there is a distinction between IQ and g. In many settings one has to be specific about which test was used, so as to talk about correlations with, say, scholastic or occupational achievements in a very specific way: this well-known test predicts this particular outcome to this extent. Useful for educators and employers.
      In research papers g is used more often. For that audience one says: the underlying intelligence level correlates with health or psychopathy or whatever, almost regardless of which intelligence measure is used.

      Delete
  6. Why is there so little questioning of the validity and reliability of classes of psychometric tasks where an average but trained person can perform at say 50 standard deviations above the mean for untrained people?

    Also, again, is anyone interested in taking a real intelligence test?

    ReplyDelete
    Replies
    1. 50 sd above the mean? What sort of psychometric tasks? I know a few people can extend their digit spans, but what else?

      Delete
    2. do most psychologists know how stupid they are?

      ---bgi participant.

      Delete
  7. How many generations does regression to the mean last?

    For example, if you want to predict a person's IQ, it's more useful to know their parents IQ racial background and their IQs than to just know their IQs. Is it much more useful to know their grandparent's race and IQs than to just know their IQs? Great-grandparents? Great-great-grandparents?

    ReplyDelete
    Replies
    1. I imagined that regression to the mean happens for each generation, because the favourable combination of genes which makes a person bright cannot be passed on without some shuffling of the cards. On balance, bright parents will have brighter children, with some drift down back to the mean. However, others have argued that it is a one generation effect. I cannot understand that. Check out Razib Khan and also Greg Cochran on this topic.

      Delete
    2. Michael A. Woodley24 April 2014 at 12:28

      Regression to the mean is one of the most widely misunderstood phenomena in all of psychology - put simply it is purely a statistical artifact, which has absolutely no substantive effect on psychometric traits across generations. It results from oversampling positive errors of measurement. In Woodley and Figueredo (2013) we illustrate the fallacy of regression to the mean as follows:

      "It needs to be pointed out that the expression regression to the mean has been historically misused with great frequency. Mathematically, regression to the mean is considered to be a statistical artifact produced by measurement error (Upton & Cook, 2006), and not a genetically substantive process occurring in the real world. For example, when one selects individuals with unusually high scores on any given measure, one also inadvertently oversamples scores with positive errors of measurement. If these errors of measurement are normally distributed and have a mean of zero, as is presumed in mathematical statistics, then any subsequent measurement of the same individuals will tend to produce lower scores, and hence retest with scores that are closer to the mean. The opposite happens when one selects individuals with unusually low scores on any given measure, while still producing subsequent scores closer to the population mean upon retesting.

      Furthermore in the case of parent-offspring correlations on g, oversampling parental scores with positive errors of measurement on IQ, as by selecting those identified as high-g individuals based on high observed IQ scores for special study, will produce regression to the mean when assessing the IQ of their offspring, even if the offspring were genetically identical to the parents, given the nature of this statistical artifact. This can be confirmed by retesting the parents themselves, which is rarely done, because one will then no doubt observe regression to the mean of the parental IQ scores in the parents themselves, presumably without having undergone any genetic recombination whatsoever. The proposition that offspring are necessarily closer to the mean of the general population in their actual latent g-factor (as opposed to their observed IQ scores) is therefore a fallacy, especially under conditions of assortative mating." (p. 66-67).

      References

      Woodley, M. A., & Figueredo, A. J. (2013). Historical variability in heritable general intelligence: It's evolutionary origins and socio-cultural consequences. UK, Buckingham: The University of Buckingham Press.

      Upton, G., & Cook, I. (2006). Oxford dictionary of statistics. Oxford: Oxford University Press.

      Delete
    3. I don't think that's correct. Regression to the mean when comparing parents and offspring is not just measurement error. Rather, it consists of all those effects that the parents cannot transmit to their offspring. Generally, only additive genetic effects and shared environments can be transmitted, while non-additive genetic effects and non-shared environmental effects (incl. measurement error) cannot.

      For example, if one of the parents has a set of alleles that together have a bigger effect than their additive effects, offspring usually won't inherit that combination. Or if the trait value of a parent differs from what is expected based on his or her genotype because there's a degree of inherent randomness in the developmental process, this non-shared "environmental" (or noise) effect cannot be transmitted to offspring, either.

      Only if additive heritability (or shared environmentality) is 100 percent will there not be regression toward the mean.

      Delete
    4. Also, regression toward the mean is over in one generation because the offspring inherit only those effects that can be transmitted across generations and thus can transmit them to their own children.

      Delete
    5. Regression to the mean studies (are supposed to) employ true scores, calculated as: Tˆ = rXX′ (X − MX) + MX. As such, measurement error is factored out; Anonymous is more than less correct.

      Delete
    6. I would like to see if you could get Michael Woodley to give a response to this objection.

      Delete
  8. To what extent does local pathogen load determine IQ?

    ReplyDelete
  9. Several papers have looked at this, with a new one coming out which I am about to review. It is a possible contributor to low ability in very poor environments.

    ReplyDelete
  10. My question: Is race a completely social construct?

    ReplyDelete