Human Diversity

by Charles Murray

  3. I’ve greatly understated the number of rare variants by using Phase 1 of the 1000 Genomes Project, which was not intended to identify a complete inventory of variants with a minor allele frequency less than .01. More on this in chapter 9.

  4. Wright (1943).

  5. Malécot (1948).

  6. Kimura and Weiss (1964).

  7. Wilson (1975).

  8. Glazko and Nei (2003); McHenry (2012).

  9. Zhu, Dennell, Huang et al. (2018).

  10. Other Homo species are placed at dates contemporaneous with or following Homo erectus. For example, Homo heidelbergensis has been thought by some to have been intermediate between Homo erectus and both the Neanderthals and Homo sapiens. Stewart and Stringer (2012); Papagianni and Morse (2015). But a controversy continues about whether Homo heidelbergensis is appropriately treated as a separate hominin. Bermúdez de Castro, Martinón-Torres, Rosell et al. (2016).

  11. This wording for the definition of AMH is adapted from Gamble, Gowlett, and Dunbar (2014): Table 1.1. The specification of “globular brain case and other traits” comes from Reich (2018).

  12. For a fascinating personal account of the struggle to figure out the origins of Homo sapiens, see a long conversation with Christopher Stringer, one of the leading protagonists. John Brockman, “Rethinking ‘Out of Africa’: Conversation with Christopher Stringer,” Edge, November 12, 2011. www.edge.org.

  13. Weidenreich (1946).

  14. Coon (1962).

  15. These dates, from Poznik, Henn, Yee et al. (2013), are the most recent estimate as I write.

  16. Stringer and Andrews (1988). Schiffels and Durbin (2014) use an innovative method to estimate genetic divergence that leads them to conclude that separation of today’s non-Africans from those who remained in Africa began around 150,000 years ago. If that is correct, and it is also correct that the dispersal that led to today’s humans occurred 40,000–80,000 years ago, the implication is that those who left Africa were already distinct from other Africans while still living in Africa. But this analysis has yet to be confirmed as I write.

  17. Cavalli-Sforza, Menozzi, and Piazza (1994).

  18. Ramachandran, Deshpande, Roseman et al. (2005): 15942.

  19. The following account was pieced together from the various sources that are cited subsequently. The most authoritative summary of the state of knowledge as I write is Nielsen, Akey, Jakobsson et al. (2017).

  20. Richter, Grün, Joannes-Boyau et al. (2017); Callaway (2017).

  21. Scerri, Thomas, Manica et al. (2018): 1.

  22. Hershkovitz, Weber, Quam et al. (2018); Posth, Wißing, Kitagawa et al. (2017).

  23. Lopez, van Dorp, and Hallenthal (2016).

  24. Nielsen, Akey, Jakobsson et al. (2017); Henn, Botigue, Peischl et al. (2016); Schiffels and Durbin (2014); Timmermann and Friedrich (2016); and Seguin-Orlando, Korneliussen, Sikora et al. (2014).

  25. For a summary of the various models for the dispersal of humans, see Groucutt, Petraglia, Bailey et al. (2015): Table 1 and accompanying text.

  26. The results of the study were presented in three papers: Mallick, Li, Lipson et al. (2016); Malaspinas, Westaway, Muller et al. (2016); and Pagani, Lawson, Jagoda et al. (2016). A commentary, Tucci and Akey (2016), summarizes them. The DNA analyses for the large databases when this study was done had sequenced each region of the genome only a few times, meaning that some errors go undetected and some SNPs are missed.

  27. Günther and Jakobsson (2016); Nielsen, Akey, Jakobsson et al. (2017).

  28. Nielsen, Akey, Jakobsson et al. (2017).

  29. Posth, Wißing, Kitagawa et al. (2017).

  30. Other evidence suggests Neanderthal introgression as early as 100,000 years ago, but those encounters apparently did not contribute to modern humans. Lopez, van Dorp, and Hallenthal (2016).

  31. Browning, Browning, Zhou et al. (2018).

  32. Brown, Sutikna, Morwood et al. (2004) and Argue, Donlon, Groves et al. (2006).

  33. Fregel, Méndez, Bokbot et al. (2018).

  34. Reich (2018): chapter 7.

  35. Reich (2018): chapter 3.

  36. Reich (2018) is the book-length account. Bae, Douka, and Petraglia (2017) provides a concise overview with emphasis on the peopling of Asia. For an account of the one-wave evidence, see Tucci and Akey (2016). For an overview of admixture with archaic hominins, see Wolf and Akey (2018).

  37. Pritchard, Stephens, and Donnelly (2000).

  38. For a discussion of the pitfalls of overinterpreting cluster analyses, see Lawson, van Dorp, and Falush (2018).

  39. Bowcock, Ruiz-Linares, Tomfohrde et al. (1994).

  40. Calafell, Shuster, Speed et al. (1998).

  41. Pritchard, Stephens, and Donnelly (2000). The current version of the program they developed, Structure, is available for download at no cost from the Pritchard Lab’s website at Stanford, web.stanford.edu/group/pritchardlab. Structure is a model-based method. The method used by Bowcock, Ruiz-Linares, Tomfohrde et al. (1994) is known as “distance-based,” and produces useful graphical representations but poses interpretive problems that Structure avoids.

  42. Rosenberg, Pritchard, Weber et al. (2002).

  43. Li, Absher, Tang et al. (2008). The Li study used a new software package similar to Structure called Frappe. The authors used the same sample as the 2002 study (reduced to 51 subpopulations and 938 cases for technical reasons). As before, no information about the subpopulations played into the software’s algorithms or “knowledge.” The program was allowed to run for 10,000 iterations, with pre-specified cluster numbers, from K = 2 to 7. Herráez, Bauchet, Tang et al. (2009) used the same database and found essentially the same results.

  44. This interpretation of ancestry coefficients, expressed as percentage of ancestry, follows the usage in Li, Absher, Tang et al. (2008) and in the article that fully explains the Frappe software. The measure is created by a vector of scores. Each score “corresponds to the probability that a randomly sampled allele from individual i originates from a specific ancestral population, k.” Tang, Peng, Wang et al. (2005): 290.

  45. Christoph Meiners, the physical anthropologist who named Caucasians in 1785, thought people from the Caucasus to be unusually handsome. The label is arbitrary, but it had to be. Try to think of any descriptive term that would cover everyone indigenous to the variety of regions where Caucasians are found.

  46. The plot is based on the value of a statistic, FST, discussed in note 4 of the introduction to Part II. It measures population differentiation due to genetic structure for each pair of individuals in the sample. It is one of the most widely used statistics in population genetics. The team conducted a principal component factor analysis on the FST matrix. Such an analysis takes a set of correlated observations—in this case, the 938-by-938 matrix of FST values—and creates a smaller number of uncorrelated variables, which are called the “principal components” (PCs). Colloquially: The program’s algorithms create a first component that explains as much of the variation among the observations as possible. It then creates a second component that explains as much of the remaining variation as it can, subject to the restriction that the second component must be uncorrelated with the first component—and so on through successive PCs until all of the variation has been explained. In the Li study, the first and second PCs explained 52 percent and 28 percent of the FST respectively, for a total of 80 percent. The first PC primarily described the contrast between sub-Saharan Africans and non-Africans, and the second PC primarily described the differences among the populations in Eurasia.

  47. Xing, Watkins, Shlien et al. (2010): 202.

  48. The conclusion reads in full:

  In this study, by sampling populations from previously undersampled regions, we sought to assess the effect of more even sampling on human genetic diversity and to investigate the evolutionary history of these populations. We found support for a relationship between the initial founding populations of America and Central/North Asian populations. We demonstrated high ge
netic diversity in Central Asian and South Asian populations, especially in Nepal. We also found that Iraqi Kurds have a closer relationship to European populations than Asian populations. These results increase our understanding of human population relationships and evolutionary history. In addition, our data provide a resource for understanding patterns of linkage disequilibrium, natural selection and the differential distributions of SNP and CNV alleles among populations, all of which have important implications in genome-wide association studies and the identification of loci with functional, biomedical significance. (Xing, Watkins, Shlien et al. (2010): 209).

  49. Novembre and Peter (2016): Fig 2.

  50. Schraiber and Akey (2015): Table 1.

  51. Schraiber and Akey (2015).

  52. An excellent review of the state of knowledge as of 2014 is Elhaik, Tatarinova, Chebotarev et al. (2014), which also describes the Geographic Population Structure algorithm that successfully placed 83 percent of individuals from a worldwide selection into their country of origin.

  53. Razib Khan, “How to Look at Population Structure,” Gene Expression, October 3, 2016, gnxp.com.

  54. Tang, Quertermous, Rodriguez et al. (2005).

  55. Long, Li, and Healy (2009) conduct an analysis based on nucleotide diversity, defined as the number of differences per site between two copies of a locus, and gene diversity, defined as the probability that two randomly drawn copies of a locus differ in state (are different alleles). Using these variables, the authors concluded that “the clustering methods in popular use produce human population groups that have a simpler structure than even the TLIM.… This structure is clearly a weak description of the true human population structure, because it does not capture the complete nested arrangement of populations.” (p. 33). This conclusion is surely true. Genetic diversity across African subpopulations as measured with these variables is greater than the genetic diversity among non-African populations, for the same reason that the genetic diversity of the East Asians left behind in the thought experiment is greater than the genetic diversity on the spaceship crew. Campbell and Tishkoff (2008). But that fact does not engage the clustering phenomenon that separates the populations in the cluster analyses (and would separate the populations in the spaceship thought experiment).

  56. Bolnick (2008); Feldman and Lewontin (2008); Feldman (2010). How has Richard Lewontin himself responded to the cluster analyses? In 2006, twelve years after Bowcock, Ruiz-Linares, Tomfohrde et al. (1994) and four years after Rosenberg, Pritchard, Weber et al. (2002), the Social Science Research Council posted a web forum under the title Is Race “Real”? One of the contributors was Richard Lewontin. “A clustering of populations that does correspond to classical continental ‘races’ can be achieved,” he wrote, “by using a special class of non-functional DNA, microsatellites. By selecting among microsatellites, it is possible to find a set that will cluster together African populations, European populations, and Asian populations, etc. These selected microsatellite DNA markers are not typical of genes, however, but have been chosen precisely because they are ‘maximally informative’ about group differences.” Richard C. Lewontin, “Confusions About Human Races,” Is Race “Real”?, June 7, 2006, raceandgenomics.ssrc.org.

  Two years later, Lewontin coauthored a paper with Marcus Feldman (Feldman was first author), “Race, Ancestry, and Medicine.” Feldman and Lewontin (2008). The paper includes a discussion of Rosenberg, Pritchard, Weber et al. (2002) and Rosenberg, Mahajan, Ramachandran et al. (2005), acknowledging the main results without dissent. But it does not back off from Lewontin’s 1972 position. “The repeated and consistent results on the apportionment of genetic diversity reviewed in the previous section show that the genes underlying the phenotypic differences used to assign race categories are atypical of the genome in general and are not a reliable index to the amount of genetic differentiation between groups. Thus, racial assignment loses any general biological interest.” Feldman and Lewontin (2008): 96. The article was written before the Li or Xing studies using hundreds of thousands of SNPs as the basis for the cluster analyses.

  57. Bolnick (2008).

  58. Rosenberg, Pritchard, Weber et al. (2002): 2382.

  59. Serre and Pääbo (2004).

  60. Rosenberg, Mahajan, Ramachandran et al. (2005): 660.

  61. Novembre and DiRienzo (2009).

  62. For an extended back-and-forth on this issue see Shiao, Bode, Beyer et al. (2012), which proposed an integration of clines with clusters as “clinal classes,” and the subsequent special issue of Sociological Theory, 32 (3), 2014, “A Symposium on ‘The Genomic Challenge to the Social Construction of Race.’”

  8: Evolution Since Humans Left Africa

  1. Escaramís, Docampo, and Rabionet (2015).

  2. Besenbacher, Sulem, Helgason et al. (2016).

  3. Kimura (1983).

  4. Geneticists distinguish between absolute fitness (in an individual, defined as the number of offspring left behind) and relative fitness (the ratio of the number of one’s offspring to the number produced by another).

  5. Haldane (1927).

  6. Darwin (1859): 127. Available at darwin-online.org.uk.

  7. Hardyck and Petrinovich (1977).

  8. Kodric-Brown and Brown (1987) summarize the long scholarly effort to understand why sex exists at all, given its evolutionary costs, and why gametes of different size are so universal across diploid organisms. Since then, progress has been made on many outstanding issues. See Parker (2014) and Sharp and Otto (2016).

  9. In one of the rare exceptions, the big-belly seahorse, the female deposits fertilized eggs in the male’s pouch. And that is accompanied by another rare exception: For that variety of seahorse, it’s the female who mates promiscuously and the male who is choosy. Geary (2010): chapter 3.

  10. Geary (2010): chapter 3.

  11. Author’s analysis, 2010–14 aggregated American Community Survey.

  12. For more on this, see Cochran and Harpending (2009): 42–44.

  13. Anderson and Stebbins (1954), elaborated in Arnold (1997).

  14. 1000 Genomes Project Consortium (2012). The following descriptions of the samples are taken from the website of the Coriell Institute for Medical Research, which is a repository for the 1000 Genomes population samples. Retrieved from coriell.org/1/NHGRI/Collections/1000-Genomes-Collections. The sample sizes (the number of genomes used in Phase 1 of the 1000 Genomes database) are taken from Moore, Wallace, Wolfe et al. (2013): Table 1. The final sample sizes in Phase 3 were somewhat larger, and are given in Oleksyk, Brukhin, and O’Brien (2015).

  The populations in Phase 1 that I used for the analyses were as follow:

  Luhya in Kenya (LWK): From Webuye Division of Bungoma district in western Kenya, who identified themselves as having four Luhya grandparents. Sample: 97.

  Yoruba in Nigeria (YRI): From one community in Ibadan, Nigeria, who identified themselves as having four Yoruba grandparents. Sample: 88.

  African Ancestry in USA (ASW): From the American Southwest, who self-identified as primarily African American and had four African American grandparents. Sample: 61.

  Han Chinese in Beijing, China (CHB): From persons living in the residential community at Beijing Normal University (who came from many different parts of China). Sample: 97.

  Han Chinese South (CHS): From Hunan and Fujian provinces, who identified themselves as having at least three out of four Han Chinese grandparents. Sample: 100.

  Japanese in Tokyo, Japan (JPT): From the Tokyo metropolitan area. “Because it is considered culturally insensitive in Japan to inquire specifically about a person’s ancestral origins, prospective donors were simply told that the general aim was to include samples from people whose grandparents were all from Japan.” Sample: 89.

  Tuscans in Italy (TSI): From a small town near Florence, who identified themselves as having at least three out of four Tuscan grandparents. Sample: 98.

  British from England and Scotland (GBR): From Cornwall and Kent (England), Argyll an
d Bute (a unitary authority council area in western Scotland), and Orkney, all of whom identified themselves as having all four of their grandparents born in the same rural area. Sample: 89.

  Finnish in Finland (FIN): From throughout Finland, who identified themselves as having at least three out of four grandparents born in Finland (98 percent had all four). Sample: 93.

  Northern and Western Europeans in Utah, USA: This sample is not described at the Coriell Institute website. They are described on the 1000 Genomes website as being of Northern and Western European extraction. Sample: 85.

  The three subpopulations from the Americas are omitted. As stated explicitly in the sample recruitment descriptions at the Coriell Institute website, these samples were not selected for Amerindian ancestry. They are representative of the populations of Puerto Rico and Colombia, and of Americans of Mexican parentage in Los Angeles, which in turn comprise a diverse mix of European and Amerindian subpopulations. I also omitted the Iberian sample from the European subpopulations because of its small sample size (14) in Phase 1.

  15. Sabeti, Schaffner, Fry et al. (2006): 1614.

  16. Hawks, Wang, and Cochran (2007): 3 of 5. For an elaboration of these issues by two of the coauthors, see Cochran and Harpending (2009): chapter 4.

  17. With important contributions from Rosalind Franklin. See Sayre (2000).

  18. I owe the analogy with Newton’s first law to Avise (2014): chapter 11.

  19. Fisher (1918).

  20. Akey (2009).

  21. Sabeti, Schaffner, Fry et al. (2006).

  22. Dick, Agrawal, Keller et al. (2015).

  23. Akey (2009).

  24. Plomin, McClearn, Smith et al. (1994); Plomin and Spinath (2004).

  25. Risch and Merikangas (1996).

  26. See Clarke, Anderson, Pettersson et al. (2011) for a step-by-step description of the analytic protocol.

  27. Risch and Merikangas (1996).

  28. Jannot, Ehret, and Perneger (2015).

  29. Hermisson and Pennings (2005): 2335.