Human Subspecies and Race

This is something that started last midweek, but that travels, a cold, and intermittent internet connectivity conspired to distract me from until now. Jerry Coyne was writing about human “race” and in so doing may have made a number of statements that some people disagreed with, and one of those people contacted me and asked my opinion, which I gave, etc. etc. etc. And, of course, the whole conversation got blogged HERE. The conversation in the comments is also very interesting. Go have a look.

Comments

  1. #1 Jason Antrosio
    March 6, 2012

    Hi Greg,

    Thank you for doing this and inspiring a new blog-post conversation. I’d just like to provide some more references here. I discuss these issues at my blog-post titled Race redux: What are people “tilting against”? and there’s also an interesting piece by Jonathan Marks, A rant on race and genetics.

    While I was researching these issues, I was introduced (at the suggestion of anthropologist Henry Harpending) to the work of geneticist Guido Barbujani. His 2010 paper (co-authored with Vincenza Colonna) is a very careful overview of the scientific literature from someone who has been studying human genetic diversity for a long time. It’s readable and current and answers many of the issues posed by Jerry Coyne: Human genome diversity: frequently asked questions. For a similar refutation from an evolutionary biologist, also see Continuous geographic structure is real, “discrete races” aren’t.

    Thanks again!

  2. #2 rork
    March 6, 2012

    Yeah, I’m in there. I agree with continuous. It makes sense too – genes flow, though some places more and some less, though using any particular words for regional differences probably always dumbs it down.

    But I am also trying to fight what I perceive as a common conclusion by non-genetics folks that “race is socially constructed” (or details about within group vs between group distances) to imply that there are no interesting gene differences as one moves around the landscape. I’m not doing so well on that nit-pick.

  3. #3 Greg Laden
    March 6, 2012

    I’m sure there are interesting gene effects as we move around that landscape, but for the most part the list of things that are widely believed that are genetically variable this way are things for which there is no known gene …. just an assumption that it must be there … and where consideration of non-genetic causes are systematically ignored rather than accounted for.

  4. #4 Lou Jost
    March 6, 2012

    Jason, thanks for the links relating to the Coyne posts. I just read the “Rant on race and genetics”, which harshly criticizes Coyne. I think that this criticism is based on a difference between anthropologists and most evolutionary biologists regarding the meaning of “race”. The rant states that race is not about differences between groups, it is about MEANINGFUL differences between groups. Yet I think most evolutionary biologists, if faced with variation in wing color of a butterfly from one island to another, would not hesitate to call these different color forms “races”, even if these differences had no selective advantages and were caused only by genetic drift. Deeper questions of meaning hardly enter. Maybe that partly explains Coyne’s position.

    As a population geneticist, I have been more disturbed by the misuse of genetic arguments in this debate. I really don’t know enough about the data to have a strong opinion about what to conclude, but I do see that many of the quantitative arguments mentioned in these posts are invalid. The Barbujani and Colonna article you linked to is a case in point. It bases part of its argument on the fact that a measure of population structure, Fst, is very low between human subpopulations. Fst is a highly nonlinear measure that is not a measure of differentiation between groups. They could have obtained low Fst values even if the groups belonged to different species entirely and shared no alleles, so the lowness of Fst is not a very good argument against the idea of race. Their description of Fst in the accompanying glossary also makes multiple mistakes: “Fst ranges from 0, when all subpopulations are identical, to 1, when different alleles are fixed in different subpopulations.” No, Fst can approach zero even if all subpopulations are completely distinct (no shared alleles). See the work of Hedrick or myself. And when there are many subpopulations, Fst can equal unity even if nearly all of them are fixed for the same allele. See the work of Gregorius. The kinds of arguments used by Barbujani and Colonna are common in the literature, and they are misleading.

    I explain this in my comment here:

    http://kenanmalik.wordpress.com/2012/03/04/why-both-sides-are-wrong-in-the-race-debate/#comment-1825

    I do not criticize Barbujani and Colonna’s conclusions (I don’t know the data at all), only the invalid reasoning.

  5. #5 Greg Laden
    March 6, 2012

    Yet I think most evolutionary biologists, if faced with variation in wing color of a butterfly from one island to another, would not hesitate to call these different color forms “races”,

    That simply isn’t true. Dividing species into nominal races when faced with variation is considered doing it wrong in evolutionary biology. The distinction you are making in your comment between “anthropologists” and “evolutionary biologists” is not correct at all.

    They could have obtained low Fst values even if the groups belonged to different species entirely and shared no alleles,

    Let’s have an example of that. (Not that Fst is conclusive here, but I’d love to see what the heck you are talking about here.)

  6. #6 Jim Thomerson
    March 6, 2012

    I don’t know about all evolutionary biologists. But when we first revised the killifish genus Austrofundulus in 1978, we considered seven geographically separated populations, statistically different at the populstion level, just as separate populations of a variable species, A. limnaeus, and discussed them as such. None showed enough difference to fit Mayer’s criteria for subspecies recognition. Later, with more collections, and DNA data, we recognized six of the populations as distinct species. A. limnaeus type locality is in the coastal desert east of Lake Maracaibo. One of the most distinctive populations is down south, around the Rio Misoa. Turns out that it is one end of a cline from the type locality, so we included it in A. limnaeus. The DNA and morphological differences are both clinal and, at the ends, not as great as DNA differences among the various species.

    It amuses me now much different the populations look, now that I know they are different species.

  7. #7 Lou Jost
    March 6, 2012

    Greg, You criticize my comments about Fst without bothering to look at the link I gave you, which has worked examples. Fst is not a measure of genetic divergence between groups, and does not have the properties or interpretation that most population geneticists ascribe to it. I’ll post my examples here later, since you clearly don’t believe me.

    Meanwhile I noticed another misconception in the Barbujani and Colonna article: “If Ne*m, the product of
    effective population size (Ne) and migration rate (m), is
    large, the effects of migration prevail and populations tend
    to converge genetically; if Ne*m is small, genetic drift plays the predominant evolutionary role, so that populations tend to diverge.” This is widely (maybe almost universally) believed, but it is false, as anyone can check by running simulations of the finite island model. The actual factor controlling cohesion or divergence of subpopulations is P/u where P is the relative migration rate between pairs of subpopulations, and u is the mutation rate at the locus of interest. Since it is the relative migration rate that matters, not the absolute number of migrants, differentiation can arise even across somewhat leaky barriers. See my comments and simulations under Nolan Kane’s post at:
    http://www.molecularecologist.com/2011/03/should-i-use-fst-gst-or-d-2/
    For the reasons why Fst or Gst do not measure differentiation of allele frequencies between groups, see Jost, L. (2008) Gst and its relatives do not measure differentiation, Molecular Ecology 17: 4015-4026. For more on the mathematics of diversity and differentiation, and the myths surrrounding it, stick [Jost diversity] into Google Scholar.

  8. #8 Greg Laden
    March 6, 2012

    Lou, I never said Fst, which is a distance measure, is what you seem to think I said it is.

    You do not give an example in that link.

  9. #9 Lou Jost
    March 6, 2012

    Greg, here’s an easy way to see the problem with Fst or Gst. Nei’s formula for Gst is (Ht-Hs)/Ht where Ht is total heterozygosity of the pooled groups, and Hs is the mean heterozygosity of the individual groups. Heterozygosity has a maximum value of 1.00, and Ht>Hs. Suppose genetic diversity within groups is high, so that Hs is close to 0.9. Then Ht must be greater than 0.9 but less than 1. So the numerator of Gst is less than 0.1. The denominator is between 0.9 and 1.0, so Gst will be less than about 0.11. Notice that this will be true no matter what, even if the groups share no alleles at all. When within-group genetic diversity is high, Gst or Fst approach zero, no matter whether the groups are similar or completely different.

    Conversely, lets suppose we have lots of groups, and all but one group are fixed for the same allele (so virtually all groups are genetically identical at that locus). The odd group is fixed for a different allele. So Hs=0 and Ht=some number. Then (Ht-Hs)/Ht =1.00, supposedly indicating maximum differentiation between groups, even though almost all groups are genetically identical. So Fst or Gst are not good ways to describe differentiation, though they do have legitimate uses for other tasks.

  10. #10 Greg Laden
    March 6, 2012

    Lou, I’m not quite sure why you are giving me a lesson in basic population genetics. I’m sure some people reading will appreciate it. I agree with you that Fst and Gst are not good ways to describe or define things like “races” for the reasons you suggest as well as for other reasons. One reason, to put it in lay terms, is that it is kind of like saying how far apart two countries are by measuring the difference between their respective geographic centers.

  11. #11 Lou Jost
    March 6, 2012

    Greg, contrary to your last comment, Fst is NOT a distance measure, and I did give worked examples at the first link I posted (the one to the kenanmalik blog).

    You questioned my statement that Fst could be close to zero even if the groups shared no alleles at all (including cases when the groups belonged to different species. I hope I answered your question and showed why Fst or Gst have no relevance to the question of genetic divergence between groups.

    My original comment was not aimed at anything you said but at the invalid reasoning in the Barbujani and Colonna article that Jason Antrosio (#1) cited favorably.

  12. #12 Greg Laden
    March 7, 2012

    Lou: Fst is a distance measure! I didn’t question your discussion of a near zero Fst, I asked for you to illustrate your case with an example, by which I mean, a known study of a known set of populations. I still think that would be interesting.

    I’ll have a look at Barbujani and Colonna when I can. This is a busy week in the world of politics so I’m having a hard time getting other stuff done!

  13. #13 Lou Jost
    March 7, 2012

    Greg, Fst is not a distance measure, nor is it even remotely a measure of differentiation of allele frequencies across demes.

    I am sending you my article from Mol Ecol (which is behind a paywall). That contains some real examples of Fst close to zero even though the demes consisted entirely of private alleles. But for readers here, I also just explained above that this apparently impossible situation will always happen when within-group variation (heterozygosity) is high. That math proof trumps even the real examples.

    Fst is consistently misintepreted in the literature and mis-taught in most standard texts. The nice thing about this (unlike most controversies) is that anyone can prove it to themselves by doing numerical examples with high-diversity demes.