Many apologies for the lateness of this post.
Previous parts: Part I, Part II, Part III
You may have noticed in previous parts of this series that I use the term "population" more often than "race." This is deliberate: if you want to determine if race is genetic, you can't assume that the genetic populations into which you divide individuals are synonymous with race. You have to first see if it's possible to divide individuals based on genetic characteristics and then determine if those populations correspond to what we call races.
In Part I, I said that a population is "is a distinct group of interbreeding people, often associated with geography. For instance, a population might be the group of people living in a single town." This is correct but not the full story. What the individuals of a population actually share is common ancestry. This isn't to say that we can pinpoint precisely the identities of these ancestors or that the individuals of a population are related in an appreciable fashion (by which I mean in a context that matters for social or cultural reasons). Nor does it mean that the humans of the world can be separated into more than one population. The belief that it's possible to separate people into genetic populations (for lack of a better term) is that geographic isolation and the issue of travel made it more likely that certain groups of people would breed amongst themselves, leading to allele frequencies being skewed in some direction or another. Natural selection might also work to skew allele frequencies. The question being answered by Witherspoon et al.'s study (not to mention many other studies) is whether or not there has been enough non-random breeding that certain populations have common ancestors that make them distinct from other populations.
With that in mind, let's break down the results that Witherspoon et al. got.
1. It's possible to separate groups of people into populations.
This is the obvious one. If you look at enough loci, it is possible to separate individuals into distinct populations. Furthermore, it is possible to separate individuals so that most individuals only belonged to one population.
2. Some individuals belonged to more than population.
Witherspoon et al. talked about individuals who had been misclassified because the methods for separating individuals placed them in a different population than what was expected. You may recall that the "expected" population was tied to the individual's race. If you look at the data, what you find is that the individual wasn't classified into the wrong population so much as the individual was a fringe case who could have potentially belonged to more than one population. The sorting method arbitrarily put the individual into one population or the other because the sorting method didn't allow for the possibility that the individual could belong to more than one population. If populations are caused by common ancestry, however, then an individual who gets sorted into more than one population is simply an individual who shares common ancestors with individuals from more than one population. When you consider that there was almost always individuals who could not be sorted into a single population in groups of people living very close to each other (where, presumably, breeding between groups could occur), this should come as no surprise.
3. The study looked at hundreds of loci, many of which have no obvious effect.
In order to make the "misclassfication" rate as low as possible, Witherspoon et al. often looked at hundreds of different loci. Many of those loci (I even venture to say most, though I can't confirm that) have no physical effect whatsoever. I don't want to get into excessive detail about basic biology, but it relies on the idea that any differences in the DNA is inheritable, whether it has a physical effect or not. Changing a single base in the DNA sequence may have no effect whatsoever (because of the redundant nature of the DNA code, or because it's in an area which doesn't code for anything) but it's still a change that can be called a different allele. It's kind of like the difference between "colour" and "color." The spelling is slightly different, but they mean the same thing.
4. Populations are not divided by the absence or presence of a few traits but, rather, by frequencies of alleles.
This is tied closely to 3. If it took simply the presence or absence of a few traits/alleles to differentiate populations, Witherspoon et al. would not have had to look at the frequency of hundreds of alleles in order to separate populations which were very far away from each other. This means that it is nigh on impossible to find an allele which is only every found in one population and never found in another population. It may be true that 99% of the individuals of Population A will have allele X while only 1% of Population B will have allele X, but allele X still persists in Population B and its existence isn't sufficient to sort an individual into one population or another. This also ties to the idea that sometimes you can find individuals between populations who are more similar than individuals within a population.
So, now we know lots of populations. The next question is whether those populations correspond to races. Since Witherspoon et al. could classify individuals into the expected populations defined by race, it would be fair to say that there is a correlation between race and genetic population. It would also be fair, however, to say that the correlation is not perfect. If we're just talking about genetic populations, then "misclassification" isn't an issue. It was, however, in Witherspoon et al.'s study, since the individuals were identified as part of a distinct race, and some could be classified into the wrong race. It could, of course, be argued that these individuals were "mixed-race" and were not misclassified at all - they were simply misidentified from the start. That would make the correlation between populations and races very strong, wouldn't it?
Of course, if genetic populations and races are the same thing, then it would mean that all the things we learned about populations applies to races. Races would be defined by hundreds of differences which have no discernible effect whatsoever on an individual's actual physical characteristics; the differences would be based on shared ancestry. The presence or absence of a few traits would be insufficient to classify an individual into a particular race, and no trait "belongs" to one race or another - they would be present in all races. Individuals between races could be more similar than individuals within the same race.
That actually doesn't sound too bad to me. The problem, of course, is that the question "Is race genetic?" isn't actually about finding out of people can genetically be split into populations defined by allele frequencies. It's about whether or not we can find a biological basis for concretely separating people into different groups which fit with the ways we've already decided to use to separate people. We want to be able to look at a person and say, "He belongs to race A because he has blue skin. She belongs to race B because she has orange skin." Instead, the data shows us that all we can do is say, "Well, he probably belongs to race A because there are more people with blue skin there. And she probably belongs to race B because there are more people with orange skin there. And, actually, they could both belong to race C. Or maybe skin colour has nothing to do with it. Let me check a hundred more loci...."
Race may correlate with genetic population, but a genetic population is not a race. Race transcends the biological, involving culture, language, upbringing, place of origin, and a host of other factors. I'd even say that these factors are even more important the biological, mainly because the biological says that we're really not all that different.
1. Witherspoon, DJ, Wooding, S, Rogers, AR, Marchani, EE, Watkins, WS, Batzer MA, Jorde, LB. 2007. Genetic similarities within and between human populations. Genetics 176: 351-359.