I've talked abut race in the past, and I'm not into the topic at this point since it is going over old ground, but a few quick responses....
Re: Lewontin's point about 85% within group vs. 15% between group variance, that is true, at one locus, but it ignores the correlation structure across loci. This is elucidated by mathematical geneticist Anthony Edwards in his paper from a few years back (PDF here). You can also think of it in terms of intersections of loci or traits. Consider that I have brown skin. If I tell you just that, you can't infer much, there are many populations which have brown skin. If I told you my hair was straight, that would narrow the possibilities some. If I told you I lacked an epicanthic fold, that would narrow the possibilies further. If I told you I was of blood group A that would narrow the possibilities further (this might seem strange to you, but when you note that my complexion overlaps with many indigenous peoples of the New World, but that these populations are overwhelmingly blood type O, with a small minority of A centered around a few tribes in northwest North America, you can use it to exclude some possibilities). Particular traits are correlated together in South Asians, and that shows. The same principle operates with genetic loci. If you want an appeal to authority, here is Richard Dawkins on page 408-409 in The Ancestor's Tale:
In short, I think Edwards is right and Lewontin, not for the first time, wrong. Lewontin did his sums right of course: he is a brilliant mathematical geneticist. The proportion of the total variation in the human species that falls into the racial partition of variation is, indeed, low. But because the between-race variation, however low a percentage of the total variation, is correlated, it is informative in ways that can be surely be demonstrated by measuring the inter-observer concordance of judgement
Developmental geneticist and evolutionary biologist Armand Leroi has been making the same point as Edwards recently, and Richard Lewontin responds here.
Another point is that there are all sorts of confusions regarding race because of sociological conditioning. Yes, this is true. I don't know what more I can say to that, genetic structure can only be approximately gleaned, and various factors like drift and selection can skew perceptions of ancestral quanta (if the latter is what strikes your fancy). Race is not just about genetics, it is about sociology, cognitive psychology and history. And I hold that it has a reality in all these domains (how that reality is relevant to how we order our societies is a different issue altogether).
Third, there are a few specific issues with John's post. He says:
The human species (convention makes me want to type "human race") is massively interbreeding.
How do you define massively? In Brazil peoples from various geographical "races" (just replace with "subpopulation" or something if you prefer) admix. In China this does not occur. Of course alleles spread, but they spread via selection as much as neutral processes, and that selection is often contingent on local conditions.
There are haplotype groups in many species, including humans, and in some species, such as the California seal, this shows geography fairly well. But not in humans. We move about too much.
Not all loci are created equal. On loci that control for skin color you can find disjoint sets of alleles. The reasons for this are obvious: local selection constrains. And here:
They rightly observe that while there are continental differences in genetics, there is no hard division, and genetic variation doesn't match up with cultural differences per se. There is a genetic substructure to the human population, but it isn't racial.
OK, let's call it genetic substructural? :) Seriously, I think a lot of this is semantic. You can abandon the word race for all I care, but ignorance of human genetic substructure can be problematic. Consider the issue of organ donation, many human populations share HLA variants, in fact, some of the variants are shared with chimpanzees. But the combinations can be rather rare, and this is why if you can't find a match from a relative looking in your "racial group" is the most fruitful tack. This is why children who are biracial have problems, they exhibit rare novel combinations far more often than typical,1 so children in the USA who are mixed-European-Thai sometimes travel to Thailand to find a match becasue this combination is more common there (some would argue that MHC diversity and structure is more important than mtDNA diversity since the former has important immunological relevance).
Within-species groupings are, evolutionarily speaking, ephemeral. Ten thousand years ago, almost none of the non-African races existed. Ten thousand years from now, almost none of the modern races will continue to exist, I warrant. And Africa is so genetically diverse (being the source of all genetic variation that hasn't evolved in the past 60,000 years) that one cannot fairly call "African" (or Ethiopian) a single group.
First, I don't agree that "ten thousand years ago" there weren't any non-African races. Europe was settled for about 40,000 years by anatomically modern humans, Australia for 50,000 years, west and south Asia for somewhat longer, and the New World had just been settled. There has been a lot of time for human variation to evolve. Additionally, John refers to Templeton's idea that there were multiple Out-of-Africa events, and admixture might have occurred. This implies that some of the alleles that float in human populations might be ancient local variants, so human diversity might be piggy-backing on an enormous alternative evolutionary history of an archaic cousin species. Also, the idea that Africa has more genetic variation needs to be qualified because that isn't true on many important loci. For example, Europeans exhibit polymorphism on MC1R to the point where nearly 30 alleles can be detected at non-trivial levels within the population. Africans tend to be constrained to the "consensus sequence" which is probably human ancestral.
The key problem with Lewontin's idea of intragroup vs. intergroup variance is that the public has some rather strange ideas in regards to this. I've been told multiple times that one is more likely to be more genetically similar to someone of another race than someone of the same race because of this enormous intragroup variance. That's not true. You can call it whatever you want, but there is non-trivial human genetic substructure on functionally relevant loci. If that's too much of a mouthful for you, just make sure that finite tissue matching resources are appropriately targeted and in that case let's not pretend like the welter of intragroup variance blinds us to the relevance of intergroup variance.
Addendum: Straight up, I think the "reality" of race is comparable to species, the difference is of degree, not kind. "Species concepts" are rough and ready, my view is one of instrumentalism, the same with "race." The magnitude might differ (at least in mammals, "species" are really problematic in many plants), but the vector is about the same.
Update: Matt McIntosh has updated the original post in direct response to John's entry. I think he gets across his point well, this is another hillist vs. mounainist between reasonable people (ie., is genetic substructure significant? Can/should we call it race?). Also, let me add one thing. Those who say race does not exist often like to point out that, there could be a
1) Lactose tolerant race (across expanses of western Eurasia and small parts of Africa)
2) Straight hair race (outside of Africa and parts of Melanesia)
3) Medium brown skin race (outside of Europe and much of the Middle East, Africa and northeast Asia)
But, if you combine these three features there is only one part of the world that really fits the bill, northern India.2 Populations are characterized by a combination of characters, not one character (though some populations can be characterized by fixation at one locus for an allele that is not present in outside populations).
1 - The MHC loci have many variants each, so can find the outcomes by multiplying the number of alleles at each loci by the number of loci like so: (10 alleles) X (7 alleles) X (8 alleles). Of course, the likelihoods are not equal, and not all populations have the same alleles, but the point is that there is a lot of allelic overlap if you look at the distribution of one allele, but the combinations tend to be much more population typical.
2 - There are Middle Eastern populations who are as swarthy as Indians, but, they tend to be characterized by a high level of African admixture, ergo, their hair is often not "straight." Additionally, African admixture would probably have driven down lactose tolerance since the Africans being enslaved are usually not the milk drinking Nilotic herdsmen. Straight haired Middle Eastern peoples on the other hand tend to have brunette white skin.
I think I just realized that if this is true:
"I've been told multiple times that one is more likely to be more genetically similar to someone of another race than some of the same race because of this enormous intragroup variance."
...then (me being white) Mos Def and Al Sharpton are more likely related to the oppressor (white slave-owners) than I am.
Something tells me that this isn't gonna fly over at okayplayer.
Re: massive intermixing in Brazil -- from what I've seen, Afro-Brazilian girls just look better on avg than Af-Ams or Africans, even if you control for the source (ordinary magazine, porn video, etc.) Probably due to recent selection, since although they might look different from other African-origin peoples due to Portuguese admixture, white desire for blacks had to start somewhere, when there was no significant admixture.
If the modern Brazilian obsession w/ female physical perfection is reliable back to 1500, and if the selection pressure is strong (like emphasizing more how hot a girl is if there are lots of bugs and you want a healthy kid), you could quickly produce some pretty atypical African-origin people. Mix in some Mediterranean features for good measure, and there you go: black girls whom white guys aren't almost universally averse to. (Ditto for black Dominicans, Puerto Ricans, Cubans...)
I think avg personality diffs factor in as well -- again, from what I hear, Afro-Brazilian girls are on avg more agreeable and less confrontational than Af-Am girls. Recent selection there isn't as obvious... I dunno, I guess if the family structure is low paternal investment, males wouldn't care so much about whether the wife was a headbutting nag or not, whereas Catholic Portuguese guys invest much more as fathers, and thus would choose the more acquiescent ones.
(for completeness: high pathogen load doesn't select for super-hot babes in the s-S homeland since women break their backs working and are chosen more for how hard-working they are, rather than how hot of a trophy wife they'd make.)
My definition: Race is a social construct that correlates with a set of apparent morphological features and/or ancestry. (Which, of course, generally correlates with genome as well.)
I think this definition reflects, pretty accurately, the one in use today by the average person.
But race isn't really a social construct. If I was completely accepted as an African-American tomorrow, this would do absolutely nothing to change my genetic affinity with Europeans.
For that matter African Amercans, Haitians, and Nigerians do not share any sort of social identity, they share a genetic identity.
In fact if American social labels like 'black' and 'white' were comnpletely dropped tomorrow, this would do nothing to change what a biologist or a sociologist mean by race, so how is the idea of 'social construct' even helpful or useful at all.
(For that matter why aren't tigers and lions just 'social constructs' that correlate with a set of apparent morphological features and/or ancestry? If we wanted to, it's not hard to say they are.)
The social construct talk is appropriate for describing some of the peculiarities in the way distinctions are drawn.
Dawkins used one like it in the ancestors tale.
Two of those people are "black" and two are "white", but that's probably not how a Martian would assign the labels, and it may be a good guess that Powell has as much European as African ancestry.
jason kidd, who seems to about about 1/4 black, has been described as a "white" basketball player in eastern europe.