In response to my post Mixed-race but homogeneous appearance? several individuals mooted the possibility that admixture may result in the vanishing of race as a social construct. Actually, I don't think this is the true. To the left is a photo from my post Can you tell if you're black or white? where I explored the genetics of a case where two black-white biracial parents produced fraternal twin daughters of disparate appearance. While one sister seemed to favor her African ancestors in look, another sister seemed to resemble her European forebears. Across the full sample space of their genome it seems likely that both these girls are about half European and half African in ancestry, but on the finite loci which are salient in the production of the features which we use to code for "races" they most certainly favor one ancestral group over the other. The basic point is that a population, and individuals, can exhibit great admixture and yet still realize the full phenotypic range of the ancestral types. This is because genetics is not blending, admixture will not result in a homogenization toward a mean unbounded by a distribution characterized by variance.
But we don't need to stick to theory or projection from sensationalized pedigrees. There is a "natural experiment," so to speak, and that is Latin America. In particular, there are two papers out of a Brazilian group which are relevant to the point I'm trying to make, Color and genomic ancestry in Brazilians & Color and genomic ancestry in Brazilians: a study with forensic microsatellites. To the left I've mashed up two figures which illustrate the most important result, and have added some text since the resolution is wanting in clarity. Here is the group's general conclusion: ancestrally informative markers tend to suggest that individuals who identify with the three broad physical groupings, white, black and mixed, exhibit a great deal less separation in terms of the variation of total genome content then one might infer from appearance. Both papers are open access, you can check to see if their methods are up to snuff. The figures show pretty clearly that there is a great deal of intersection across the phenotypic races. Granted, a group which is avowedly mixed-race would naturally overlap the ancestral clusters, but do note that the black and white groups also overlap quite a bit, though the extent of overlap exhibits the patter one would expect. The authors placed next to the Brazilian clusters controls for European and African ancestry, which show that these clusters much more closely match our expectation that physical appearance should track ancestry. Whenever I post on this topic I do get very skeptical Brazilian readers, and I don't dismiss objections as to the representativeness of the sample, but these studies are what I've found so far. I would not be surprised to find that on the national level the whites of the far south of Brazil are overwhelmingly European, for example. Additionally, someone who is "white" in Brazil may exhibit features which mark them of some African ancestry in other contexts. That being said, I am willing to bet that assortative mating driven by social selection has resulted in the reemergence of phenotypic races from a admixed population over several generations across large swaths of Brazil. In plain English, imagine a scenario where the "white" twin above marries another "white" twin, and the "black" twin marries another "black" twin. Though the cousins produced by these pairings will all be about 1/2 African and 1/2 European in ancestry, in physical appearance they will resemble two distinct races. Though their genetic distance using ancestrally informative markers would place them between their African and European ancestors, the "white" cousins would no doubt be the beneficiaries of positive discrimination and the "black" ones negative discrimination. And that, I believe, is the key as to why the assortative mating occurred in the first place, in much of Latin America a great deal of physical variation exists within families and there is a distinct preference for the offspring with "good features," invariably European features. These genealogically mixed-race individuals who are phenotypically white would naturally tend to solidify their lucky draw by finding a partner with the same qualities and so pass on to their children the benefits of being white. There is the idea in the minds of some that interracial relations in the most intimate sense will banish racism or racial consciousness. To me that seems unlikely, though Latin American nations do not exhibit the sort of racism based on ancestry traditional in the United States, they do retain a marked preference for those who wear a European face, no matter what their ancestry might be. The reality of multiracial origins certainly softens the hard edge of bigotry, but it also seems to encourage a complacency as to blatant phenotypic biases. Selection is a powerful force and it can reshape human variation rather quickly given the appropriate environmental impetus. The race problem will not vanish through the fiat of genetics, rather, the social context can have a controlling effect upon the correlation of characters and their clusters within the population so as to reflect the values which that society holds up.