First, some levels of genetic clustering corresponded nicely with certain folk racial classifications. For example, at K = 3, the clusters picked out Mongloids, Caucasoids, and Negroids. Furthermore, at K = 5, the clusters picked out current U.S. Census races: Sub-Saharan Africans (hereafter ‘‘Black Africans’’), Caucasians, East Asians, Amerindians, and Oceanians. K = 6 was similar to K = 5, except it had a single Pakistani population (the Kalash) in a part all by itself. A second interesting result was that the level of populations with the highest among-part genetic variance (r2a ) was K = 5, which was 4.3 ± 0.40 at 95% confidence (Rosenberg et al., 2002, p. 2382). This last result seemed to support Risch et al.’s famous claim that ‘‘genetic differentiation is greatest when defined on a continental basis’’ (Risch, Burchard, Ziv, & Tang, 2002, p. 3). Despite these promising results for racial naturalism, Hochman argues that the genetic clusters of human populations identified in Rosenberg et al. (2002, 2005), and in other similar genetic studies, do not satisfy any of (a)–(d).
First, citing Ning Yu, Chen, Ota, and Li (2002), Hochman (2013, pp. 336, 349) points out that Black Africans house the most genetic diversity out of any ‘‘continental population’’. In fact, according to Yu et al. (2002), Black African subpopulations are more genetically different than Black Africans and Eurasians. Specifically, Yu et al. (2002, p. 269) found that the average nucleotide diversity (p) for 50 autosomal noncoding DNA segments (with 146 single nucleotide
polymorphisms, or ‘‘SNPs’’) in 10 Black Africans (from 9 populations) was 0.115 ± 0.016%, while p for the same DNA segments and SNPs between the same sample of Black Africans and 20 Eurasians (from 20 populations) was only 0.096 ± 0.012%. The calculations were repeated using other autosomal and X-linked regions (Yu et al., 2002, p. 273). The result appeared to be robust that
there is more genetic diversity in Black Africans than there is genetic distance between Black Africans and Eurasians. But this only shows that genetic clusters of human populations do not satisfy half of (a). Hochman (2013, p. 348) uses genetic clustering results from Sarah Tishkoff et al. (2009) to show that there is more genetic structure among Black African populations than among populations at any level of genetic clustering that Black Africans arise as a genetic cluster.
But before I discuss Tishkoff et al.’s results, I should stop and quickly clarify something. It is important to note that ‘genetic diversity’, ‘genetic distance’, and ‘genetic structure’ are noninterchangeable technical terms for Hochman that (roughly) correspond to what Jonathan Kaplan and Rasmus Winther (2013) have called ‘‘genetic diversity’’, ‘‘genetic differentiation’’, and ‘‘genetic heterozygosity’’, respectively, which are three distinct ways of measuring genetic variation. According to Kaplan and Winther (2013, p. 404), genetic diversity is ‘‘a measure of how heterogeneous a system is [genetically].’’ So, for example, p would be a good
example of a measure of genetic diversity.5 In contrast, genetic differentiation
is ‘‘a measure of how different two things are [genetically],’’ and is exemplified by genetic distance calculations, such as Fst genetic distance (Kaplan &Winther, 2013, p. 404). Finally, genetic heterozygosity is ‘‘a measure of the fraction of heterozygotes (as opposed to homozygotes) in a population,’’ and is exemplified by Sewell Wright’s F-statistics or the U-statistics of AMOVA (Kaplan & Winther, 2013, p. 405). With that clarification behind us, we can return to Tishkoff et al.’s results. Tishkoff et al. (2009) used Structure, principle component analysis (or ‘‘PCA’’),6 1327 genetic markers, 3945 individuals, and 181 human populations from all across the globe. Using Tishkoff et al.’s results, Hochman points out that Africans have an incredible amount of genetic structure. Tishkoff et al. (2009) identified 14 genetic clusters of African populations (including Afro-Americans). Just as interesting is that when looking at the worldwide sample, Tishkoff et al.
(2009, p. 1038) found that at KP 4, Black Africans split into different
lusters! For example, at K = 5, the clusters were Caucasoids, Mongloids, Western and Central Africans (including Afro-Americans), Eastern Africans, and the Hadza people of Tanzania.7 Just from looking at these results, it is no wonder why Hochman concluded that genetic clusters of human populations do not satisfy (a)…
As for (a), genetic distance (or structure) need not be greater among two or more clusters of populations than genetic diversity (or structure) is within any cluster of the set because, again, this condition need not hold in order to have population structure. In fact, Yu et al.’s results are exactly what we should expect for Homo sapiens. Since our species has spent much more time in Africa (P140,000 years) than it has outside of Africa (660,000 years), our species has had a long time to develop a considerable amount of genetic diversity and genetic structure in Africa (Cavalli-Sforza & Feldman, 2003, p. 270). Given this historical reality, we should not expect to find as much genetic diversity or genetic structure outside Africa as we find inside Africa. This, of course, does not imply that humans do not have population structure along (roughly) continental boundaries, since Black Africans, on average, could still be more likely to mate with other Black Africans than with non-(Black Africans). With that said, Tishkoff et al.’s results are concerning because we should expect B to arise at K = 5 if B is a human population partition.
However, once one examines Tishkoff et al.’s sampling scheme, one will find that Tishkoff et al.’s results are non-threatening to Rosenberg et al.’s results. This is because Tishkoff et al. (2009) dramatically oversample African ethnic groups. African ethnic groups make up 65.1% of Tishkoff et al.’s sample, even though African ethnic groups make up just 30.2% of human ethnic groups.16 Combined with the known fact that most K = 5 human genetic variation lies within parts, not among them, it is unsurprising that a study that greatly oversamples one major geographic region will find that its ethnic groups split into different genetic clusters at K = 5. In fact, Tishkoff et al.’s result is not unique. Friedlaender et al. (2008) sample human populations in such a way that Oceanian ethnic groups make up 48.6% of the sample, even though Oceanian ethnic groups make up just 18.5% of human ethnic groups.17 At K = 5, Friedlaender et al. (2008, p. 178) find that the partition consists of Caucasians, non-Oceanian Mongloids, Black Africans, and two distinct clusters of Oceanians! Of course, neither Tishkoff et al.’s nor Friedlaender et al.’s results conflict with the standard result that K = 5 human genetic clusters are Amerindians, Black Africans, Caucasians, East Asians, and Oceanians, because their samples of human ethnic groups are not appropriate for identifying worldwide human population structure. Rather, these samples are
ppropriate for studying African population structure and Oceanian population structure, respectively. (Spencer, Q. (2014). The unnatural racial naturalism. Studies in History and Philosophy of Science Part C: Studies in History and Philosophy of Biological and Biomedical Sciences,.