does race exist? part trois
This semester we spent a lot of time discussing Shiao et al’s (2012) article in Sociological Theory claiming that recent genetic research provides a reason to believe that “races” exist. Now we’ll discuss the symposium that was recently published.
There are three responses by Anne Morning, Daniel Hosag, and Fujimora et al. There is a lot in there so I will focus on what I think is most important:
- Genomic analyses are contaminated by racial categories. I.e., genomic clustering results rely on social categories of race.
- Genomic analyses are inconsistent in that different algorithms producs different number of human clusters (“clinal groups”).
- Genomic analyses do not clearly map onto groups that would clearly be identitified as racial or ethnic groups.
Let me take each in turn. On a purely logical level, 1 is probably the weakest point. As I noted in my original post on Shiao et al 2012, the contamination of a scientific research program by social bias does not logically imply that the basic idea is flawed. That requires a different argument. My original example: social definitions of non-humans have plagued scientific research, but that doesn’t imply that there aren’t meaningful distinctions between fish or rocks. It only shows that a particular scientist got it wrong.
Point 2 is a much stronger point. Inconsistent results, or those that are very sensitive to initial paramters set during model estimation, should reduce our confidence. I think the respondents do a good job suggesting that genomic research does not show a clear partition of people based on genomic data.
I think Shiao (and other people on his side) have a plausible response: human populations have no clear boundaries, they intermix a bit, and we should expect fuzzy boundaries. To support this point, you would examine the distribution of the number of clusters done using different data and different methods. If we get a very “flat” posterior (i.e., any number of groups is equally possible), the critics win. If the distribution has concentrated mass and the mean is not zero, then Shiao et al wins. In other words, meta analysis is the way we settle this sort of claim. Neither side has done this analysis in the Sociological Theory exchange.
Point 3 is unpersuasive as presented. As I noted in an earlier point, it is logically possible that there is genuine clustering of people but it doesn’t match to our notions of what counts as a race. For example, maybe I am not Latino but I am Basque-Dutch-Colombian-Sub-Tico. So race exists, but not in the way we understand it. So the mismatch between genomic data and folk notions of race may be beside the point.
Shiao et al’s response hits some common points and focuses on others (e.g., their review of the genomic literature is accurate in contrast to what the critics claim). Shiao et al’s response to point 2 is a bit different in that it goes into detail about what certain algorithms accomplish.
Overall, I am struck at what was accepted by most folks. There seem to be genuine biological differences between people, behavioral genetics is not irrelevant to sociology, and there seems to be meaningful dimensions of variation among people that is tied to geography. This last point is also noted by Shiao at el. Shiao then makes a strong point – if you believe that there is genomic variation by geography, why immediately jump to the strongest constructionist argument? Doesn’t make sense.
A few months ago, I noted that I was a racial agnostic because I don’t possess the technical knowledge to judge rival claims and I don’t immediately assume the constructionist view is true. After reading this exchange, I am moving toward the view that there is indeed systematic variation in people, but “Race” might be a terribly bad way to think about it.