Population-genetic trees,maps,and narratives of the great human diasporas

Blood group studies: The genetic remaking of human groups, human evolution as a Brownian motion, and the first attempts at a global dendritic order

One of the points of origin of contemporary human population genetics consists in the research on the distribution of the human blood groups that goes back to the First World War. After the Second World War, blood group studies came of age. Among the centers that bound together a research network were Arthur Mourant’s serological laboratories in Britain. ² As Jenny Bangham (2013) shows in her PhD thesis, Mourant’s collecting, analysis and publication of genetic data were also part of a wider network between colonial administrations, metropolitan labs, blood banks, anthropological studies, and institutions such as transfusion depots, colonial hospitals, public health centers and university departments. Scientists turned to Mourant for identification of sampling populations and sampling techniques; they asked him for experts on particular subjects; they told him about their population genetic projects and expeditions; they asked for anti-sera or sent him their blood samples to be analysed in his lab; and most importantly, they drew on the blood-group data that Mourant and his team centralized in tables and books – theirs was also the largest collection of offprints containing such data in the world.

While for many scientists the blood group population studies were a more objective technique to define human races and their interrelations, from the 1930s onwards, they were also seen as a tool against the old notions of racial types. Mourant was among those who regarded the study of blood groups as a ‘scientific anthropology’ (Mourant, 1954: 1) that could replace the anthropological racialism of old. Nonetheless, his effort to centralize and standardize the blood group data cemented rather than undermined the practice of reifying existing categorizations. This becomes obvious when looking at Cavalli-Sforza’s interactions with him. In the early 1950s, Cavalli-Sforza, who was then at the Institute of Serotherapy in Milan, wanted to set about realizing his old plan of a project to develop new statistical methods for the analysis of the genetics of human populations and enquired after blood group frequency data. While at the beginning of the 1960s he relied on Mourant in a project on a region in Italy with regard to anti-sera, in 1963 he informed him of a global study with Antony Edwards to which I return below (Wellcome Library, London, Sforza, L. L. Cavalli, shelfmark: PP/AEM/K.120: box 33, reference number: b17736328, correspondence 1953–63). Another 10 years later, Cavalli-Sforza wrote to Mourant because he and Edwards wanted to work on a greater number of blood system markers but had not found all the data needed in the literature. Figure 5 is the table by way of which he asked Mourant for the missing information, which the latter provided on the sheet. While Mourant, for example, could not help with regard to the genetic serum protein marker Gc (group specific component) in ‘Basques’, he could point to the literature for data on Tf (transferrin) typing for this group, and he even provided some data. With regard to ‘W. Germany’, to give another example, he noted down average frequencies for the Tf alleles. He also provided Cavalli-Sforza with tabulations that he had made for the new edition of his monumental The Distribution of the Human Blood Groups (1954, 1976). As Figure 5 illustrates, the genetic data were associated with a range of population labels from regional and national to ethnic and ‘aboriginal’ (Figure 5).

OPEN IN VIEWER

Figure 5.

Mourant to Cavalli-Sforza, 1 October 1973, Wellcome Library, London, Sforza, L. L. Cavalli, shelfmark: PP/AEM/K.121: box 33, reference number: b17736894, correspondence 1965–73.

The above observations also point to the fact that geneticists might work with data produced by others, rather than themselves collecting and typing blood. In these cases, the identification of populations was done by someone else, and the inconsistent use of population labels in the literature could render the work difficult. Furthermore, the sampling of ‘remote’ populations who were of particular interest to Cavalli-Sforza had only recently become feasible on any larger scale due to, as Joanna Radin (2013) has shown, the pre-war industrial production of refrigeration machines and the post-war knowledge of how to prevent damage to cells in freezing and defrosting. Therefore, populations and sample individuals might have been chosen for their easy accessibility apart from their relative genetic isolation. I will return to this aspect in the context of the Jewish model and the first sequence trees. But first I have to engage with the early visualizations from the blood system data.

At the beginning of the 1960s Cavalli-Sforza became professor of genetics at the University of Pavia, and, with the data within reach, he invited the British statistician and geneticist Anthony Edwards to work on the informatization and quantification of phylogenetic studies. Edwards had taught himself to work with computers at Cambridge, and the Italian Ministry of Education, too, invested in computers that appeared on the market in the early 1960s. The University of Pavia received a room-sized Olivetti Elea 6001. On the basis of this infrastructure, Cavalli-Sforza and Edwards developed statistical methods to build evolutionary trees, and Edwards wrote computer programs for them (least-squares based on an additive tree; minimum evolution/maximum parsimony [Edwards]; and maximum likelihood after Fisher). For the method of maximum likelihood that compares statistical hypotheses for their likelihood (which is proportional to the probability of the data given the hypothesis), Edwards wrote EVOTREE. Both men had their statistics from Fisher’s Statistical Methods for Research Workers (1925), and for Edwards, they aimed at the ‘embedding of modern phylogenetic analysis in the Fisherian tradition of stochastic model building and efficient statistical estimation by maximum likelihood and the use of genetical data’ (Edwards, 2009: 10). However, even though EVOTREE would soon be overtaken by other programs, in the beginning it was rather the other two solutions – distance matrix and parsimony – that were taken up by the community. The more complicated maximum likelihood required more time. ³

It was on the basis of the method of parsimony, or the assumption of the least evolutionary change to account for the data, that Cavalli-Sforza and Edwards built their first populational tree. With the help of Mourant, they analysed published data on 20 alleles related to the 5 major blood group systems from 15 human populations (3 per continent). And they liked to emphasize that this resulted in ‘the first evolutionary tree of human populations’ (Cavalli-Sforza, 2009: 8; see also 1992) (Figure 6). We again find a wild mix of population labels, some of which are also present on the sheet shown in Figure 5. Even before publishing the tree as shown in Figure 6, however, Cavalli-Sforza and Edwards incorporated it (without the population labels) into another global mapping technique: that of projecting a phylogenetic tree based on human genetics onto a world map. At this stage, the way they projected the as yet unpublished tree onto a map resulted in a sweep of the lines that does not suggest a journey out from Africa. One might read the picture as placing the cradle of humankind in the area of today’s Iran (Figure 7). Furthermore, the main split in this early tree existed between ‘Europeans’/‘Africans’ and ‘Asiatics’ (Cavalli-Sforza and Edwards, 1965: 929). We are thus clearly in the pre-revolution period of molecular anthropology, before the consensus tree and narrative of the out-of-Africa model had emerged. The archeological and mitochondrial-DNA evidence of the 1980s for the common African root would undermine this early tree and map. The consensus on the general scheme of the human story was strongly driven by the revolution in human population genetics through DNA-sequence analysis. And it was in fact a paper coauthored by Cavalli-Sforza in 1983 that showed that the ‘Caucasian’ samples clustered more closely with those labeled ‘Oriental’ and ‘Am. Indian’ than with the ‘Bantu’ or ‘Bushmen’ (Johnson et al., 1983: 267) – a development I look at in more detail below.

OPEN IN VIEWER

Figure 6.

The evolutionary tree of human populations from blood group polymorphism frequencies produced by parsimony, L. L. Cavalli-Sforza and A. W. F. Edwards, ‘Analysis of Human Evolution’, in S. J. Geerts (ed.) Genetics Today: Proceedings of the XI. International Congress of Genetics, The Hague, The Netherlands, September 1963, vol. 3 (Oxford: Pergamon, 1965), pp. 923–33 (p. 929).

OPEN IN VIEWER

Figure 7.

A.W. F. Edwards and L. L. Cavalli-Sforza, ‘Reconstruction of Evolutionary Trees’, in V. E. Heywood and J. McNeill (eds) Phenetic and Phylogenetic Classification (London: the Systematics Association, 1964), pp. 67–76 (p. 75); with kind permission from the Systematics Association.

The early trees and the ways in which they rendered human kinship and history differed from the later more popular ones based on mitochondrial DNA. First, they were built from classical markers, or those polymorphisms (different alleles) that had been discovered by immunological and electrophoretic analysis of gene products such as proteins, and not on the differences in the nucleotide sequence of the DNA. Second, these trees did not trace the genealogies of DNA sequences (the historical sequence of individual mutations), but were based on the average genetic differences between entire sampling populations. In the simplest form, the genetic distance between two populations was measured by the difference in the frequency of a gene. The assumption was that the genetic difference between two populations was on average proportional to the time that had elapsed since their separation (genetic distance); but there was also a spatial dimension in that this genetic distance was taken to be proportional to distance in space. Thus, the trees resulting from the genetic data taken from human groups globally were meant to approximate a history of successive population splits related to the geographical distribution of actual living populations across the world. The tree diagram here literally is a map (Deleuze, 1988[1986]: 44). Several assumptions had to be made: each population evolves independently; the cause of divergence between populations is mainly random genetic drift; and populations split at random and into two daughter populations identical to the parent. However, this linearity might have been tempered, for example, when a small group split from a population and became isolated through migration. This would have meant that it carried with it a sample of genes that was not representative of the mother population. Because of such possibilities, the difference between the populations under comparison had to be established as the average of the differences in frequency for a large number of genes (the law of large numbers). ⁴

Overall, evolution was thus conceived as ‘a branching random walk’ with a constant probability of branching and a constant rate of walking (Cavalli-Sforza and Edwards, 1967: 256). And it is not irrelevant that during this time, Cavalli-Sforza and Edwards also made the personal acquaintance of the Japanese population geneticist Motoo Kimura, who played an essential role in the development of the neutral mutation theory. In 1965, Cavalli-Sforza invited Kimura to Pavia, and in 1968 the latter published his theory that the speed of evolution at the molecular level was determined by the rate of mutation. This was attributed to the fact that mutations mostly have neither a positive nor a negative selective effect for their carriers; they were described as selectively neutral and as spreading randomly through a population (Kimura, 1968; Dietrich, 1994). The theory was in line with Cavalli-Sforza’s and his colleagues’ findings from the Parma valley concerning the role of drift in isolated populations. The fact that the differences in frequencies between villages were similar for different alleles suggested that the frequencies were the result of drift rather than selection (Cavalli-Sforza, 1966). This research they would continue and eventually summarize in Consanguinity, Inbreeding, and Drift in Italy (Cavalli-Sforza, Moroni and Zei, 2004). On the basis of these assumptions, Cavalli-Sforza and Edwards developed tools for building human phylogenetic trees. The complex realities of human intermixture and of human agency more generally were lost. What remained were trees that reduced human history to ‘a branching Brownian-motion process’ (Cavalli-Sforza and Edwards, 1967: 256).

Cavalli-Sforza was well aware of the many problems associated with constructing phylogenetic trees from genetic data, and – reminiscent of Huxley’s cautioning – he even suggested that it might work only for populations that are geographically far apart, because otherwise ‘[i]nstead of a “tree” one may have to estimate a “network”; such methods do not yet exist’ (Cavalli-Sforza, 1973: 96). At that point, he also emphasized that the distance analysis underlying phylogenetic trees differs from our understanding of kinship. While kinship denotes genetic similarity by descent or inbreeding, genetic distance measures genetic difference due to nature (or chance). He was furthermore cautious about the linearity between kinship and geographical distance as well as between genetic difference and distance in time (for example, the possibility to assume regular average mutation rates [Cavalli-Sforza, 1973: 90–1]; for a later treatment see, for example, also Mountain and Cavalli-Sforza, 1994). However, as Steve Olson has pointed out: ‘Cavalli wrote about the limitations of using phylogenetic trees in depicting population histories, but his decision to use them anyway shaped the subsequent history of the field’ (Olson, 2006: 172). Cavalli-Sforza was certainly troubled by some of the limitations of tree-building, and, in particular, models that could take into account or check for the factor of admixture were needed. If the human expansion was to be reconstructed from existing populations, it was important to know whether the assumption of the persistence of a genetic diasporic signal in presumably isolated populations was justified.

Jewish population genetics as a model for reconstructing the human diaspora

At the beginning of the 1970s, Cavalli-Sforza accepted a professorship at Stanford University, where he built up one of the major laboratories for human population genetics. A few years later he informed Mourant that he and his colleague Dorit Carmelli were ready to analyse some Jewish data. They were preparing a computer program and wanted to test it on a small sample. They wrote that

… the history of the Jewish Diaspora brings to our attention some interesting experiments of nature bearing on some basic tenets of human population genetics. A population, relatively homogeneous at first, gave rise to a number of splinter groups which occupy many different niches in many different places. (Cavalli-Sforza and Carmelli, 1979: 93)

In other words, the Jewish diaspora could function as a model to approach questions with regard to the reconstruction of modern human evolution at large. What was the relative importance of drift versus natural selection in the explanation of present gene frequencies? To what degree had there been exchange of genes between diaspora and ‘surrounding’ populations? Jewish communities were assumed to be relatively isolated. They also suggested themselves for studies because they were often within easy reach, and there were historical records and traditions to test hypotheses (see also Lipphardt, 2010: 310).

Extensive genetic data from different Jewish groups on the distribution of classical genetic markers such as blood groups, enzymes, serum proteins, and the immunoglobulin and Human Leukocyte Antigen systems had been accumulated since the 1950s. However, while Mourant had already presented a huge amount of data on the Jewish communities in his The Distribution of the Human Blood Groups of 1954, it amounted to a confusing cacophony. He complained that ‘[a] study of the ABO frequencies of Jews living in different parts of Europe, Asia and Africa shows that these frequencies vary widely from one place to another. The Jews are not a relatively homogeneous population like the more recently dispersed Gipsies and Armenians’ (Mourant, 1954: 70). It was only in the late 1970s that population genetics produced a more coherent picture. Cavalli-Sforza and Carmelli’s work was part of this achievement. They wanted to analyse several Jewish/non-Jewish population pairs to get at the unknown fraction of genes that each Jewish group had received from the donor population in its home country. Because there were several donor populations for every Jewish population (such as Ashkenazim), and the Jewish populations had also migrated, the gene frequencies for these donor populations would have to be averaged out. In this way, Cavalli-Sforza wrote in his letter to Mourant, ‘one can hope to estimate the gene frequencies of the original Jewish population, the amount of admixture each has undergone with local populations, the extent of drift and deviations that may have to be accounted for by selection’ (Cavalli-Sforza to Mourant, 25 September 1974, Wellcome Library, London, Sforza, L. L. Cavalli, shelfmark: PP/AEM/K.122: box 33, reference number: b17740186, 1973–74).

Thus, while Cavalli-Sforza and Edwards had not yet been able to estimate hybridization for the 1965 tree, the Jewish case study allowed experimenting with models. Eventually, Cavalli-Sforza wanted to see whether the reconstructed ‘original’ Jewish gene frequencies matched with those obtainable from the Middle East. He wanted to know whether the current inhabitants could stand in for the original ones. He therefore asked Mourant to suggest population pairs beyond those he had already found in Mourant’s publications, which Mourant did, also sending him copies of the proofs for the new edition of The Distribution of the Human Blood Groups (1976) and other Jewish data (Wellcome Library, London, Sforza, L. L. Cavalli, shelfmark: PP/AEM/K.122: box 33, reference number: b17740186, 1973–74). The results of Cavalli-Sforza and Carmelli’s study obviously depended on how a Jewish group was defined. As Nadia Abu El-Haj (2012: 87) has shown, in the early blood group studies in Israel in the 1950s and 1960s, Ashkenazim and Sephardim were taken as a priori groups to be sampled as unities, while Near Eastern and North African groups were treated as individual cities, nations, or regions. This was reflected in Mourant’s data and thus in the study by Cavalli-Sforza and Carmelli, which also affected the donor populations considered. ⁵ For example, to estimate admixture for Ashkenazim, the gene frequencies were compared with Central Europeans (in contrast, some newer studies suggest that Ashkenazi samples cluster with southern Europeans). Finally, despite the fact that the paper was about several Jewish communities, the general results of the study were published in 1979 under the title ‘The Ashkenazi Gene Pool: Interpretations’.

Thus, Cavalli-Sforza and Carmelli estimated the original general Jewish gene frequencies from the Jewish and non-Jewish groups under concern (maximum likelihood), and they found ‘that they lie rather close to the center of the Near East cluster’ (Cavalli-Sforza and Carmelli, 1979: 96). In other words, they had justified the practice of taking a living population as proxy for an ancestral group. In fact, further down in the paper they use ‘the average of the Middle East data as a substitute for the original Jewish gene frequencies’ (ibid.: 98). With regard to admixture they concluded: ‘We find that admixture with local populations was probably small in most cases. In Ashkenazim, however, total admixture may have been of the order of 40%, but there is significant heterogeneity between different markers’ (ibid.: 101). They explained that if each Jewish population were to be put in one of the 4 host groups (Central Europe; Southern Europe; North Africa; and the Near East), most would be classified with the area from which they originally came (the Middle East) rather than with the area in which they had lived since the diaspora. Thus, they wrote that ‘[p]lotting on the same graph the individual Jewish populations, one notes considerable scatter [that does, however] not cancel the possibility of recognizing the Middle-Eastern origin of the Jews’ (ibid.: 97). Finally, the scatter was apparently random, and Cavalli-Sforza and Carmelli concluded that the genetic changes that had taken place in the Jewish diaspora were most likely due to drift.

What this paper was meant to establish was a common origin of Jewish diaspora groups in the Near East and that the diaspora groups were genetically still closer to the populations of their geographic origin than to their host populations. In fact, in The Genetics of the Jews of 1978, Mourant brought forward the entire amount of serological data his group had collected from people and publications, charted and tabulated, in support of these conclusions of the Cavalli-Sforza/Carmelli paper, the results of which he had received before publication (Mourant, Kopeć and Domaniewska-Sobczak, 1978).

Getting at the sequences: The mitochondrial tree and the consensus story of human expansion

The 1980s saw something like a revolution in human population genetics. It was possible to sequence DNA and especially mitochondrial DNA was becoming a favored epistemic object among other reasons because it is only maternally transferred – it seemed to come without the complexities of recombination (Sommer, 2008). Cavalli-Sforza and collaborators wrote that ‘[s]ome of these studies have indicated the existence of significant differences in frequencies of mtDNA between racial groups and have also suggested that shared mtDNA polymorphisms may indicate group affinities [10] within major ethnic groups’ (Bonné-Tamir et al., 1986: 342). In fact, two important studies had come out of the Stanford lab. In the first, the scientists had interpreted the distribution of mtDNA types as pointing to Asia as a very important site in the radiation of human ethnic groups; they wrote that

… [t]he suggestion for these mtDNA data that the human ethnic groups diverged from an Asian origin is consistent with comparable data on nuclear gene frequencies. These results suggest that formation of human ethnic groups took place in the last part of the Pleistocene, starting in Asia and radiating towards the other continents, probably within the last 50,000–100,000 years … (Denaro et al., 1981: 5771)

It was the second paper that Cavalli-Sforza co-authored that signaled the arrival at a broad human origins consensus. The team published the first population tree based on mtDNA in 1983 (Johnson et al., 1983). It was built on the analysis of 200 individuals of 5 different populations. After digesting the DNA with restriction endonuclease, the fragments were separated on horizontal agarose slab gels and visualized through autoradiography. The digestion with different enzymes provided 32 Restriction Fragment Lengths Polymorphism patterns, or morphs, which combined to 35 mtDNA types. The researchers found that the major ethnic groups exhibited quantitative and qualitative differences in types. When represented in a parsimonious tree (i.e. assuming the minimum number of necessary mutations to account for the data), closely related mtDNA clustered according to geographic regions (see Figure 8). The three central types showed high frequency, appeared in more than one ethnic group, or were present in primates. The different mtDNA types were shown to be related by steps of mutations and the reconstruction of this sequence was understood to provide insights into the biological history of humankind.

OPEN IN VIEWER

Figure 8.

M. J. Johnson, D. C. Wallace, S. D. Ferris, M. C. Rattazzi and L. L. Cavalli-Sforza, ‘Radiation of Human Mitochondria DNA Types Analyzed by Restriction Endonuclease Cleavage Patterns’, Journal of Molecular Evolution 19(3–4) (1983): 255–71 (264), Figure 6 (with kind permission from Springer Science and Business Media).

The scientists also used the morph frequencies to produce an average linkage tree (see Figure 9). When equal evolution rates were assumed for all branches, the tree separated the ‘Bushmen’ first from all others. If they estimated the location of the root from the three central types, both ‘Bushmen’ and ‘Bantu’ were on one branch and the rest on another, with the ‘Bushmen’ showing a significantly higher mutation rate. Finally, the mtDNA trees also came with dates. Using an immunological timing of the human-gorilla/chimpanzee split (5–7 million years), the authors arrived at an estimate of human ethnic diversification of approximately 100,000 years ago. This date was within the range of estimates calculated from protein markers by Cavalli-Sforza and others in the 1960s, but it was gained under the assumption that all the ethnic groups studied had diverged simultaneously. However, when instead following what they took to be the more likely pattern of diversification shown in Figure 9, the dates arrived at for the first split (‘Bushmen’ at approx. 220,000 years) and for the last one (‘Oriental’/‘Am. Indian’ at approx. 5,500 years) were out of the range of archeological evidence – a fact that again suggested unequal mutation rates in the various ethnic lines. Notwithstanding these difficulties, the authors could conclude that there was a high correlation between mtDNA type and the ethnic origin of an individual; and they had planted the root of the family tree in Africa (Johnson et al., 1983).

OPEN IN VIEWER

Figure 9.

M. J. Johnson, D. C. Wallace, S. D. Ferris, M. C. Rattazzi and L. L. Cavalli-Sforza, ‘Radiation of Human Mitochondria DNA Types Analyzed by Restriction Endonuclease Cleavage Patterns’, Journal of Molecular Evolution 19(3–4) (1983): 255–71 (267), Figure 7 (with kind permission from Springer Science and Business Media).

The new technologies were also used to improve the Jewish model. There had been one particular hitch in the Cavalli-Sforza and Carmelli (1979) paper. They had worked with non-Jewish samples from the Near East. In other words, what this analysis suggested was that the communities of the Jewish diaspora would have to be classified with the non-Jewish Near Eastern samples. This seems to have been supported by the fact that some of the Jewish Near Eastern groups showed admixture rates of close to 100% with their host population (which the authors took to be non-representative). It remained to be shown that the Jewish people could not only be isolated from their host populations in the diaspora but also from the Near Eastern gene pool. In the research that went into the paper ‘Human Mitochondrial DNA Types in Two Israeli Populations – A Comparative Study at the DNA Level’, a team of scientists that included Cavalli-Sforza now wanted ‘to introduce this new tool [of mtDNA analysis] to human genetic studies of the diverse ethnic communities now gathered in Israel, and it is further hoped that polymorphisms in mtDNA will yield an independent measure of the degree of relatedness between different communities’ (Bonné-Tamir et al., 1986: 342).

Although they spoke of the comparison between two distinct Israeli communities, 41 samples stemmed from Arabs picked at random from several villages in central and northern Israel and 40 from randomly picked Israeli-born Jews at Stanford, comprising 35 of Ashkenazi origin, 3 Sephardi, 2 Yemenite and 1 Iraqi [Bonné-Tamir et al., sic]. The resulting Restriction Fragment Lengths Polymorphism patterns were compared between Israeli Arabs and Jews and to results from ‘Caucasian’ (50), African (40 Bantu) and ‘Oriental’ samples (46). Using 5 enzymes, 8 new morphs were observed. Although both groups were understood to belong to the same major ethnic division (‘Caucasians’), they differed significantly from the existing Caucasian data in the fragment patterns resulting from certain restriction enzymes. Of the 18 different combinations of fragment patterns found (combination in each individual to form an mtDNA type), only three were shared by the Jewish and non-Jewish groups, but with striking frequency differences, and the Arab sample showed ‘“African” characteristics’ (i.e. typical African markers and types; Bonné-Tamir et al., 1986: 341).

These facts were to speak for the distinctiveness of Israeli Jews both from the non-Jewish population in their region of origin and in the diaspora. The authors concluded that

… [i]n spite of the very small numbers in each group, these data demonstrate the possible existence of group-specific mtDNA fragment patterns and that certain types may be unique to certain groups. Even in a relatively short evolutionary time new types appear and may remain unique in populations that are relatively isolated one from the other. (Bonné-Tamir et al., 1986: 348)

Apart from the fact that a relatively small sample of American, mainly Ashkenazi, Jews came to stand in for Israeli Jews, in this kind of research, a community such as the Israeli Jews becomes identified with (frequencies in) DNA-fragment patterns (Figure 10). Again, the genetics of the Jewish diaspora suggested that the modern human diasporas could be reconstructed on the basis of the comparative study of ‘isolated populations’.

OPEN IN VIEWER

Figure 10.

B. Bonné-Tamir, M. J. Johnson, A. Natali, D. C. Wallace and L. L. Cavalli-Sforza ‘Human Mitochondrial DNA Types in Two Israeli Populations – A Comparative Study at the DNA Level’, American Journal of Human Genetics 38(3) (1986): 341–51 (345) (© Elsevier, with kind permission from Elsevier).

The 10 new mtDNA types were also integrated in the tree of 1983 (Figure 8) through single site changes. More than half of the new patterns radiated directly from the central and most common type 1, two from ‘Caucasian’ types, and one from an African lineage (Figure 11). Despite this success, the tree was still far from elegant. In fact, where such mtDNA studies were concerned, Cavalli-Sforza’s team was eventually outrun by Allan Wilson’s lab at Berkeley, out of which had already come spectacular immunological primate trees. In the HGDP, the two would join forces, but in the 1980s, Cavalli-Sforza felt scooped by the Wilson team (Stone and Lurquin, 2005: 139–51). They used a more efficient mtDNA technique that had been pioneered by Wes Brown, who had been brought to Berkeley by Wilson (see, for example, Brown, George and Wilson, 1979; Brown, 1980). The Wilson team also carried out a more extensive analysis. Indeed, the results coming out of the Berkeley lab caused an earthquake that reverberated far beyond the anthropological communities. Rebecca Cann, Mark Stoneking and Wilson sequenced mtDNA of approximately 150 people from what were called African, Asian, Aboriginal Australian, Caucasian and New Guinean races or populations (Restriction Fragment Length Polymorphism maps). The results of the mtDNA study were published in an article in Nature that contained a tree based on mtDNA types (Figure 12).

OPEN IN VIEWER

Figure 11.

B. Bonné-Tamir, M. J. Johnson, A. Natali, D. C. Wallace and L. L. Cavalli-Sforza, ‘Human Mitochondrial DNA Types in Two Israeli Populations – A Comparative Study at the DNA Level’, American Journal of Human Genetics 38(3) (1986): 341–51 (350) (©Elsevier, with kind permission from Elsevier).

OPEN IN VIEWER

Figure 12.

Reprinted by permission from Macmillan Publishers Ltd: R. L. Cann, M. Stoneking and A. C. Wilson, ‘Mitochondrial DNA and Human Evolution’, Nature 325(6099) (1987): 31–6 (34) (with kind permission from Nature Publishing Group [http://www.nature.com/]).

Again, what Cann, Stoneking and Wilson (1987) published was a computer-built diagram of sequence differences in mtDNA molecules, based on maximum parsimony and midpoint rooting (middle of longest distance between two mtDNA types). In the tree, the people sampled, and the races and regions they were taken to be representatives of, are reduced to DNA sequences. In the text, we still find Osborn’s and Keith’s racial groups that in the tree denote geographic regions. Furthermore, African Americans could stand in for sub-Saharan Africans because the authors reasoned from their understanding of the history of slavery that African Americans would have Caucasian nuclear genes from white fathers but not Caucasian mtDNA from white mothers. They made the unstated assumption that white men had children with black women but not vice versa. The tree is thus racialized and gendered in intricate ways. At the same time, the tree makes clear that there are ‘[m]ultiple lineages per race’ (ibid.: 33). In fact, although, or possibly because it is more refined, the Cann, Stoneking and Wilson tree is messier than the Johnson et al. tree of 1983 shown in Figure 8. Both of these trees do not suggest clear-cut racial groups, however, because they visualize individual mtDNA types; it is in the population trees such as the Johnson et al. tree shown in Figure 9 that the individual scatter is hidden behind population labels.

Clearly, the Cann, Stoneking and Wilson tree supported the African root, and so did the narrative that accompanied it. Beyond phylogeny, the results suggested a scenario of human evolution in the form of a diaspora that again proved to be gendered as well as racialized. It seemed that all human mtDNA referred back to a female who had lived in Africa some 140,000 to 290,000 years ago (on the basis of archeological dates for first peopling of New Guinea, Australia and the New World and regular mutation rate). This was understood to be in agreement with the archeological evidence. What came to be known as the African Eve theory or recent African evolution model was a prehistory of human origin, migration, separation and struggle reminiscent of the biblical narrative (and innumerable other tales of human quest). Modern humans had originated in Africa (that is, modern Homo sapiens evolved from the archaic in Africa), from where they began to migrate some 100,000 to 140,000 years ago to eventually spread across the globe. In the process, the modern human newcomers completely replaced archaic Homo sapiens (including Neanderthals) in Asia and Europe. The Cann, Stoneking and Wilson paper was a milestone in the establishment of a broad consensus in population genetics and archeology with regard to the picture of human evolution (Sommer, 2008). The rival notion of multiregional ‘racial’ development under considerable genetic exchange, the visualization of which Weidenreich had pioneered, became marginalized. Even though Cann, Stoneking and Wilson consciously chose the tree that linked individuals over the population tree (that would have been less parsimonious), modern human evolution had become a diaspora from Africa. And the population trees would not be a long time coming. ⁶

As we have seen, the Cann, Stoneking and Wilson paper was not the first population genetic study based on mtDNA. Because of the consensus and narrative they suggested, however, it was their story that attracted the attention of the press. According to Cann (1997), the association of the African Eve theory with the female line and with an African origin also triggered sexist and racist reactions that purportedly took issue with the science of the research:

The gang-trashing of the African ‘Eve’ or Lucky Mother can be linked to an insistence that something essential which had been previously ignored, overlooked, or misunderstood was to be learned about our history by concentrating on genes inherited exclusively from women. (Cann, 1997: 83)

In the next section, I engage with different issues that anthropologists had with the population-specific trees. Cann’s defense of the genetic approach in general seems warranted, but when she writes that ‘[w]hen phylogenetic trees based on mitochondrial molecules are becoming standard in fields of vertebrate systematics (Hillis and Moritz, 1990), the objections raised by anthropologists (and some geneticists) to this development [in anthropology] is troubling’ (Cann, 1997: 85), she is not sensitive enough with regard to the history and power of the tree and the possible confusion of trees with only diverging branches between species and such within the human species.

The great human diasporas, or the troubled journeys of population-genetic trees and maps

Cavalli-Sforza has not only engaged in synthesizing different fields of knowledge on the basis of population genetics; he has also made a great effort to communicate the insights from his science to experts in other fields, students and publics. The Great Human Diasporas: The History of Diversity and Evolution gave a personal account of his research on our ‘genetic heritage’ for a non-specialist audience (Cavalli-Sforza and Cavalli-Sforza, 1995[1993]: x). The achievement that the very influential The History and Geography of Human Genes (Cavalli-Sforza, Menozzi and Piazza, 1994) constituted was rendered more accessible to the non-expert in Genes, Peoples and Languages (Cavalli-Sforza, 2000[1996]). Through these and other writings, Cavalli-Sforza popularized a human journey that would in its essentials be retold again and again: modern humans left Africa some 100,000 years ago and began to expand across the globe via the Middle East. One branch moved along the south-eastern coast all the way to Australia (more than 40,000 years ago); another migrated through Central Asia (possibly some 60,000 years ago) and from there into southern Asia, Europe (c. 35,000 years ago) and as well into the New World (between 35,000 and 15,000 years ago). Following this general outline, in History and Geography (Cavalli-Sforza, Menozzi and Piazza, 1994), the detailed multidisciplinary histories were treated for Africa, Asia, Europe, the Americas and Australia/New Guinea/the Pacific islands. This story of the human diasporas was accompanied by trees and maps in the tradition of the first visualizations of this kind that Edwards and Cavalli-Sforza published (see Figures 6 and 7) – only by now the root and origin were clearly placed in eastern Africa (Figure 13). Despite the fact that the assumption of a ‘treeness’ of modern human genetic kinship has continued to be challenged (e.g. Templeton, 1999: 638–40 on Cavalli-Sforza), both the narrative and imagery of History and Geography occupy a core place in the history of human population genetics and its popularization. In fact, in a review of the field by Cavalli-Sforza and his long-time collaborator at Stanford, Marcus Feldman, that recapitulated the major developments in genetic history up to 2003, they reproduced the visual summary of human expansion across the globe shown in Figure 13 nearly unchanged (Figure 14 ).

OPEN IN VIEWER

Figure 13.

L. L. Cavalli-Sforza, P. Menozzi and A. Piazza, The History and Geography of Human Genes (Princeton, NJ: Princeton University Press, 1994), p. 156 (with kind permission from Princeton University Press).

OPEN IN VIEWER

Figure 14.

Reprinted by permission from Macmillan Publishers Ltd: L. L. Cavalli-Sforza and M. W. Feldman, ‘The Application of Molecular Genetic Approaches to the Study of Human Evolution’, Nature Genetics 33(3s) (2003): 266–75 (270), Figure 3 (with kind permission from Nature Publishing Group [http://www.nature.com/]).

The work on History and Geography was also closely linked to the HGDP of which Cavalli-Sforza was the main signatory on the 1991 call. Both were associated with the desire to establish a collection of cell lines from ‘isolated peoples’ worldwide and to carry out comparative analyses of DNA and other markers in order to arrive at the definitive human story, population genetic tree and global genetic map. Both heavily relied on the similarity between the genetic and linguistic trees, a parallelization that came in for criticism, not only because many linguists did not agree on the tree structure of language evolution, but also because it reinforced the impression of intra-human evolution as tree-like (Ridley, 2000). The parallelism was used to justify the identification of populations on the basis of language. Cultural characteristics were thus drawn on to identify ‘populations’ that were then genetically reified in the science. With the trees traveling into other disciplines, some of these population labels were shown to be the outcome of colonial exploitation. The population genetic trees seemed to naturalize and fix historically contingent, contested and changeable identifications (e.g. MacEachern, 2000). Reminiscent of Huxley’s warnings with regard to a priori categories in anthropological studies, the question of the best sampling method initially also troubled the HGDP. Before his sudden death, Wilson had opposed Cavalli-Sforza’s approach of sampling predefined groups with the suggestion of a grid method.

The concerns that some scientists and scholars voiced stood in conflict with the humanist discourse of the HGDP. It was presented as a pan-human enterprise that would provide people with knowledge about who they are and where they come from. The HGDP was advertised as an endeavor that would reveal the unity underlying human diversity and thereby counteract xenophobia and in particular racism. At the same time, Jenny Reardon (2001; 2005: ch. 5) has shown that the HGDP was informed by the notion that an objective scientific approach would allow access to a socio-culturally neutral kind of knowledge; and according to her analysis it was this notion that turned the well-meant initiative into a vampire project – an expression of racist and colonialist exploitation – in the eyes of the World Council of Indigenous Peoples. Reardon has argued that the belief in a politically neutral science blinded the project participants to the fact that also population genetic studies always co-produce the social and the natural order. Because of this ignorance, the research project was not designed in communication with the ‘research subjects’, so that indigenous rights organizations and some indigenous groups feared for the autonomy of First Nation Peoples. The project did not seem to be aimed at their future welfare, but only interested in the benefit to western science and economy. In the discussions about the project between experts and in the public, the issue of the scientific legacy of a racist anthropology as part of the colonial enterprise was raised. With its aim to collect bodily material from indigenous groups and to define and classify them, the HGDP seemed to stand in that tradition (on the HGDP, see also M’Charek, 2005). ⁷

This brings me back to the point made at the beginning that the technologies of tree-building and mapping that have been transformed through genetic methods are entangled with the history of European expansion and the exploitation and subjugation of communities. In the context of the HGDP, the concerns of indigenous groups were exacerbated by their conceptualization as stuck in a pre-modern past, and by the notion that genetics was superior to their ways of understanding themselves and their history. This stance was in fact clearly expressed in the document produced after the last planning workshop for the project. The report stated that ‘[s]tudy of these populations optimizes the ability to reconstruct the ethnographic map to its state at the beginning of recorded history’ (The Human Genome Diversity (HGD) Project, 1993: 4). Confusingly then, it is history that the genetic trees and maps try to eliminate from the picture. They visualize a kinship system and migration pattern before intermixture is supposed to have taken place on a greater scale. To complicate matters further, they at the same time represent genetic variation among living populations who thereby become ‘our’ past (in the sense of the past of industrialized people) (e.g. Sommer, 2008, 2010b; Reardon and TallBear, 2012).

As in Foucault’s notion, such diagrams freeze motions or submit them to rules. It is in this context that one may speak of human population genetics as a manifestation of the physics of power that according to Foucault consists in mechanisms that analyse distributions, movements, series, combinations, and use instruments to render visible, to register, to differentiate and to compare. It is structured by those disciplinary technologies that ensure the ordering of human diversity (Foucault, 1976[1975]: 268, 279–82). Among these technologies, the population genetic tree is a geopolitical diagram – it makes visible and maps on the surface of the earth the otherwise invisible underlying kinship structure of humanity (ibid.: 264). At the same time, it brings this order into being. Jeremy Bentham’s plan of the ideal prison appeared to Foucault as the diagram of the modern disciplinary society, and for Deleuze there is such a thing as ‘the diagram of feudalism’ (Deleuze, 1988[1986]: 40) in that the diagram becomes co-extensive with the entire social world (ibid.: 34). Along these lines, the human population genetic tree may be regarded as a diagram of the postcolonial world order – it embodies the legacy of colonial anthropology and the unequal power relations in place after the historical colonial periods.

Although the protest from indigenous rights organizations stopped public funding of the HGDP, it is still ongoing. Some two decades after the call for the HGDP, the HGD Panel included a collection of lymphocyte cell lines from more than 1,000 anonymized individuals, belonging to 52 ethnic groups from the 5 continents (there is no phenotypic information, only on the population from which the samples came). As one outcome, a group of Stanford researchers including Cavalli-Sforza made available on the Internet and published in Science the results of an analysis of the 948 unrelated individuals of the HGD Panel collection for 650,000 DNA nucleotides known to be most variable in humans (Li et al., 2008; see also Cavalli-Sforza, 2005). This represented the largest set of data available on the genetic variation of any species, and its full statistical analysis was estimated to occupy the researchers for some time to come. In early 2013, more than 100 labs had accessed the HGD Panel DNA.

The HGD Panel is meant to allow researchers to apply their increasingly refined methods and technologies to the same samples of human cell lines. In fact, the samples had been used for important population genetic research prior to the one just mentioned. There had been an application of Structure to the analysis of autosomal microsatellites (Rosenberg et al., 2002). This as well as the Stanford team’s study indicates that genome-wide analyses have become more common and that there have been significant developments in information technology. Programs such as Structure (developed by a team including Jonathan Pritchard [Pritchard, Stephens and Donnelly, 2000], who had received his PhD at Stanford under Feldman) and Frappe (Stanford School of Medicine, Tang et al., 2005) allow research to estimate individual ancestry proportions and to group individual genetic samples without prior determination of ethnicity. And as the figure from the paper co-authored by Cavalli-Sforza indicates, they also visualize admixture (Figure 15a). ⁸ It seems, therefore, that after all Huxley’s vision of human population genetics has come true, because already in The Science of Life, a collaborative work of Huxley and the Wellses, it was suggested that it might be possible in the future to determine the various ethnic contributions to the genetic inheritance of an individual (Wells, Huxley and Wells, 1934 [1929–1931]: 1448–9). And in We Europeans, Huxley and Haddon declared that

… [w]e can thus no longer think of common ancestry, a single original stock, as the essential badge of a ‘race’. What residuum of truth there is in this idea is purely quantitative. Two Englishmen, for instance, are almost certain to have more ancestors in common than an Englishman and a negro. For the sharply-defined qualitative notion of common ancestry we must substitute the statistical idea of the probable number of common ancestors which two members of a group may be expected to share in going back a certain period of time. Being quantitative and statistical, this concept cannot provide a sharp definition of race, nor do justice to the results of recombination. If, however, concrete values for the probability could be obtained for various groups (which would be a matter of great practical difficulty), it would provide a ‘coefficient of common ancestry’ which could serve as the only possible measure of their biological relationship. (Huxley and Haddon, 1935: 106)

OPEN IN VIEWER

Figure 15a.

From J. Z. Li, D. M. Absher, H. Tang, A. M. Southwick, A. M. Casto, S. Ramachandran, H. M. Cann, G. S. Barsh, M. Feldman, L. L. Cavalli-Sforza and R. M. Myers, ‘Worldwide Human Relationships Inferred from Genome-Wide Patterns of Variation’, Science 319(5866) (2008): 1100–4 (1101); reprinted with permission from AAAS.

Indeed, human population geneticists are increasingly interested in reconstructing the admixture events in human history. Drawing on the HGD Panel data, and on the premise ‘that human populations have interacted throughout history’, a recent paper in Science has presented results of genetic admixture that can be correlated with such historical events as ‘the Mongol empire, Arab slave trade, Bantu expansion, first millennium CE migrations in Eastern Europe, and European colonialism’ (Hellenthal et al., 2014: 747). However, Huxley and Haddon identified a difficulty with this conception. It still suggests a state in the past when there were isolated human groups, the traces of which might now be recovered in the genomes of individuals and the gene pools of populations. But such a notion of ‘primary subspecies’ they called purely hypothetical, ‘a matter of inference only’ (Huxley and Haddon, 1935: 264).

For this reason, the admixture models have more recently been criticized by multiregionalists, who continue to reject the notion of the evolution of modern human populations as having taken place in relative isolation (with subsequent admixture) and by inference the claim that the tree structure can adequately capture any point in human evolution. They maintain that the genetic data are compatible with a model that combines migration and expansion events (out of Africa) with regional descent and gene flow between regions (constrained by isolation through geographic distance) as visualized in Figure 16. ⁹ Of course, at least for the layperson, the tree does not even suggest admixture. This becomes drastically obvious in the image that represents the data of Figure 15a in the shape of a tree (Figure 15b). If the information is visualized in a tree, in which populations function as unities rather than groups of admixed individuals, we lose that mosaic picture. We end up with a diasporic structure in which populations have differentiated but not converged. ¹⁰

OPEN IN VIEWER

Figure 15b.

From J. Z. Li, D. M. Absher, H. Tang, A. M. Southwick, A. M. Casto, S. Ramachandran, H. M. Cann, G. S. Barsh, M. Feldman, L. L. Cavalli-Sforza and R. M. Myers, ‘Worldwide Human Relationships Inferred from Genome-Wide Patterns of Variation’, Science 319(5866) (2008): 1100–4 (1101); reprinted with permission from AAAS.

OPEN IN VIEWER

Figure 16.

A. R. Templeton, ‘Haplotype Trees and Modern Human Origins’, Yearbook of Physical Anthropology 48 (2005): 33–59 (50), Figure 9 (©2005 WILEY-LISS, INC.).

Concluding thoughts, or the question of the ‘treeness’ of human diversity

The tree imagery and the diasporic accounts of modern human evolution have been accompanied by a seemingly subversive discourse. Already in Genetics, Evolution, and Man (1976), Walter Bodmer and Cavalli-Sforza explained that there exist no pure races, that there has always been some admixture, that the distribution of most traits is clinal, and that frequency distributions change with traits. This last fact was driven home visually by maps of blood group distributions among ‘aboriginal populations of the world’; their confusing patterns show the messiness of the genetics of populations (Bodmer and Cavalli-Sforza, 1976: 562–74). This was not to say that the genetics is meaningless, however. Rather, unlike trees based on anthropometric data that might support our racial intuitions because they group parallel adaptations to similar climates, genetic population trees were said to be potentially counter-intuitive (ibid.: 584–7). At the same time, Bodmer and Cavalli-Sforza maintained that, on the basis of genetics, humanity may still be subdivided into three large races: the Africans, Caucasians and ‘Easterners’. They emphasized that ‘the broadest geographic groupings do correspond to a large extent with the available genetic data’ (ibid.: 574).

Two decades later, Cavalli-Sforza opened Genes, Peoples, and Languages (2000[1996]) by retelling the story of the anthropological journey from the surface of the body to the very foundation of life. He deconstructed racial prejudice and the biological basis of a particular understanding of race under the heading of genes and history. He repeated the story of a physical anthropology of race that had been based on superficial and meaningless differences such as hair type and skin color, and that in the end failed to define clear-cut races. Then came genetics that, too, did not document clear-cut and certainly no pure races. Instead of the races of physical anthropology, the nucleotide sequences revealed the secret of history. For historical reconstructions, those molecular differences that are not affected by natural selection had turned out to be more useful than those that tell how a group of humans has adapted to a particular environment. The traces of history in the gene have been left by chance. They are hidden away from the context, located in the seemingly unimportant random mutations that over fairly long periods and averaged out over data from an adequate number of genes occur reasonably regularly. The rationale thus was that while in the scientific journey from skin to DNA-sequence we lose a particular understanding of race, we gain a new understanding of history. Or, as Feldman has suggested, thinking along the lines of race should give way to an understanding of identity and diversity according to ancestry (Feldman, 2010).

However, Lisa Gannett and James Griesemer have described this shift already for the blood group work in the first half of the 20th century. They arrive at the conclusion that we are not dealing with a replacement of race by ancestry, but with a redefinition of race along genealogical rather than ecological terms (Gannett and Griesemer, 2004: 147). Cavalli-Sforza declared the characters of physical anthropology to be meaningless for a new way of grouping people according to traits that have not been under natural selection. This new grouping practice therefore runs contrary to everyday notions of race as something visible to the eye and expressed in behavior. However, as we have seen, he nonetheless worked with certain group labels of old and confirmed categories such as ‘Caucasians’ that are associated with physical traits. Last but not least, the trees show isolated populations, and they are no longer counter-intuitive. ¹¹

The fact that Richard Lewontin proved that around 85% of human genetic variation is found on the individual level and not the group level is often repeated in both expert and lay contexts to deconstruct the biological meaning of race. However, Edwards explains that this is misleading, because Lewontin only looked at loci independently. In other words, the ‘treeness’ that Cavalli-Sforza and others found in the human genetic variation is based on ‘hidden’ correlations between loci (Cavalli-Sforza and Piazza, 1975). According to this reading, the c. 15% between-populations variation is of classificatory significance, because not all the characters studied are independent. With each such locus added to the comparison, the percentage of between-population variation rises (though not in relation to the within-population variation). Edwards thus disclaims frequently made statements such as that human races are biologically meaningless, that two individuals of the same group are as different as two individuals of different groups, that we differ because individuals differ and not because we belong to different populations, and that one cannot predict somebody’s race from her or his genes. Genetic information from individuals that is subjected to cluster analysis such as Structure brings to light ‘genetic affinities that have unsurprising geographic, linguistic and cultural parallels’ (Edwards, 2003: 801). In fact, in a collaborative paper, Feldman and Lewontin (2008) have emphasized that although Lewontin’s numbers for within- and between-population variation have often been confirmed, it does not follow that individuals do not genetically cluster according to geographical regions. They also maintained the possibility to assign an individual to a particular geographical region of origin on the basis of his or her genome. ¹²

However, discursively and visually foregrounding the treeness of human diversity constitutes a choice to focus on a particular kind of variation rather than another. It constitutes a choice for dichotomous visualization and narration of human evolutionary history along the lines of a root-and-branching structure. As we have seen at the beginning of this article, this choice has been critically examined for a long time – not only with regard to humankind. In fact today, with increased awareness of horizontal genetic transfer and the recognition of intertaxa, conceptualizations of organismic phylogeny in terms of networks rather than in terms of a universal tree of life have gained in currency (e.g. Doolittle, 1999). If phylogenies of higher taxa are messy, this is certainly the case for inner-human variation. Let me give the last word to Huxley and Haddon:

The evolution of man and, specifically, of the existing varieties, is frequently represented in the form of a tree. Much misunderstanding has unfortunately arisen in the course of evolutionary discussion by the use of this term and its dependents, ‘stem’, ‘stock’, ‘branch’, ‘root’, and so on … [E]volutionary branches may – if they have not proceeded too far – unite again after they have diverged and then either rebranch or remain united. This reunion of branches is especially pronounced in man. (Huxley and Haddon, 1935: 161)