I’ve been sitting on this paper for a while, hoping I could get some geneticist to collaborate on the write-up, because while I recognise enough of the words in genetics at least not to fall off when the argument goes round corners, I certainly can’t evaluate whether it’s soundly based or not. Simon Ford back at Clare in Cambridge gave it a once-over and thought it basically sane, though—and my thanks to him—and the point has come to write about it. Please bear in mind that I am not an expert in this stuff and would welcome any corrections or different perspectives, and read on. The work in question is a paper with twenty different authors that appeared in the American Journal of Human Genetics for 2008 entitled “The Genetic Legacy of Religious Diversity and Intolerance: Paternal Lineages of Christians, Jews, and Muslims in the Iberian Peninsula”.¹ What it does is to take samples of the DNA of 1140 people from all over modern Spain and Portugal and compare them to similar datasets from Morocco and Tunisia plus an independently-derived average one for Sephardic Jews. The point of all this is the historical context that the Iberian Peninsula has, for a lot of its history, had a considerable Jewish population (`Sephardic’ actually comes from the Hebrew for Spain, ‘Sefarad’) and, of course, between about 710 and 1610, a fairly significant Muslim one that was ruling most of the area for most of that time. And, despite the fairly pompous title, what this paper does is compare what they consider to be the known history of the peninsula with the current genetic traces of population admixture.²

"Haplogroup Distributions in Iberian, North African, and Sephardic Jewish Populations." "Sectors in pie charts are colored according to haplogroup in the schematic tree to the right, and sector areas are proportional to haplogroup frequency."

And these are, in a sense, the results. You have to realise, before you start to read this thing, that we are in a game not of certainties or clear causation here but rather of statistically significant correlation. So, you will notice that the three African samples (which they got from elsewhere³) are dominated by haplogroup E3b2 (and I’m not even going to try and explain what a haplogroup is; I would just have to copy it from a better explanation like this one anyway) but the Iberian ones are dominated instead by various branches of R1. This doesn’t stop each of those groups having some trace of the other one’s dominant element, because these things occur throughout most of humanity by now, and the question is not usually down to a single genetic signature like the ‘Cohen gene’ but to a pattern being convincingly like another pattern. If you compare it to the Sephardic Jewish signature at top right (again, from elsewhere, on this occasion a separate survey by two of the authors whose data is only given as supplemental information online⁴) you’ll see that there the significant marker seems to be the balance of groups G, J2 and all other J groups, which is a bit harder to spot. So, rather than just try and spot colour matches it seems worthwhile to say what the paper’s authors think they’ve found, given that they have crunched this data in a number of other ways that don’t make such colourful images, and then remark on that. Their conclusions were, roughly:

1. Obviously, the Gibraltar Straits do mark a genuine divide in the make-up of the populations, which is not to say that there’s no common blood (ultimately, after all, we’re all cousins) but that there is a statistically significant (and fairly obvious) difference.
2. The Basque country and Gascony have a strong showing from haplogroups that barely show up elsewhere (R1b3f, otherwise only strongly represented in Catalonia and the Balearics, weirdly; R1b3d; and in the Basque Country proper, R1b3b, which actually doesn’t show up anywhere else on the plot except for a sliver in North-West Castile, although I wonder if the big sample size there might not be something to do with that) and are also statistically quite different because of that.
3. In the peninsula overall, admixture from an African-type parental population appears to be 10·6%, but this varies widely; there is none in the Basque zones, but 21·6% in Castile, i. e. twice as much as elsewhere.
4. Admixture from a Jewish-type population is rather higher, 19·8% overall, but again with variation: none at all in Minorca, but 36·3% in Southern Portugal.
5. The diversity of haplogroups within the dominant one from Africa is lower in the Iberian Peninsula than in Africa, suggesting that only a subset of the African population as it now is is represented in the peninsula’s genes.
6. Contrariwise, the diversity of the Jewish sample in the Iberian population is higher, suggesting a longer-term admixture (though see below).
7. The African sample is represented, not as one might expect most strongly in the south around Granada and least strongly in the north, but rather in the west, especially Castile and Galicia, that is the furthest parts north of the west, as well as also in Minorca which is less surprising maybe.⁵

Some of this makes perfectly good sense with what we know of the demographic history of this area, although it does persistently have to be borne in mind that we are talking about a history covering all of the last, say, three thousand years, piled up and indistinguishable. There is some possibility of distinguishing chronology with such evidence: as the authors say, the low diversity of the African sample in the Iberian peninsula compared to the Jewish or African-local ones suggests that it arrived more recently than the others because it has presumably had less time to spread and average out. But this is not ‘proving’ the Muslim conquest from genetics or anything; it is noticing a particular phenomenon that the conquest we already knew about provides an obvious explanation for. Likewise, the strong Jewish signal in South Portugal is odd until you consider that Portugal, unlike Castile or Aragon-Catalonia, didn’t expel its Jews and therefore picked up quite a lot of exile population from the reconquered areas of those two kingdoms, i. e. the south, in the fifteenth century, who have presumably left some trace in the genepool since then. On the other hand, the western-side bias of the African signal is very strange. It is certainly true that Muslim settlement, for most historians at least, is unlikely to have been substantial outside of Córdoba’s immediate zone of control, and we can do quite a lot about suggesting from place-names which groups wound up where.⁶ That would explain the low signal in Catalonia, but it patently conflicts with the high signal in Galicia. The authors suggest that this is down to the forced relocation of the morisco populations to the north and west after the war of 1567-71,⁷ and so indeed it may be—we have to watch that we don’t immediately conclude “OMG settlement in 711!” from this data given that it also includes all movements since—but if so it seems very strange to me that the areas where we know Muslims were for longest show less of a trace, and that suggests that the incomers were distinctive and also didn’t mix very much, whereas the moriscos blended into the wider population much more.

Horseshoe arches in the Leonese church of Santiago de Peñalba, another kind of evidence for cultural admixture, more or less contraindicated by the genetic evidence

There are also three problems with their conceptual framework that I see which I think need discussion. The first, they have anticipated and headed off, although they don’t phrase it quite as I would, which is that if this Jewish signal prototype they have is already based on Jews from only this area then inevitably, you’d think, it is going to be much more admixed with Iberian material than a sample of a population outside Spain. In other words, there is a risk here of concluding, “Iberian Jews… are quite Iberian“, what is somewhat less exciting than the assertion “Iberian populations surprisingly Jewish”, which is more like what the paper actually says. The authors were not worried about this, as far as I can tell, but were concerned that the self-identification of the Jewish population that was used to obtain the sample on which they relied, i. e. the DNA of people who think they’re Jewish by descent, might well be less exclusively Jewish than those people thought. The counter that they have to this also works for my worry, however, it being that there are within that Jewish sample, as well as quite a lot of haplotypes shared with Iberian populations, three or four that are not but do match strongly with the Middle East. So, as long as they aren’t Greek or Phœnician (the long time-frame again), which seems unlikely given that they are as strongly visible in the West as the East, that does show some genuine Semitic ancestry to the sample group.

"Mean North African, Sephardic Jewish, and Iberian admixture proportions among Iberian samples, based on the mY estimator and on Moroccan, Sephardic Jewish, and Basque parental populations, are represented on a map as shaded bars on bar charts. Error bars indicate standard deviations, and three-letter codes indicate populations, as given in Figure 1."

That takes us straight to the second problem, though, which is one of missing populations. I am broadly happy that most of the Muslim army of 711 and subsequent settlement was probably composed of Berbers and other Africans (even if allegedly some of them might have been Vandals by descent…), I think this is one of the things that Guichard’s work makes acceptable, but nonetheless they weren’t all Berbers, there was an actually-Arabic presence in the officer corps, because we know some of them by name, and of course there was also a massive civil war in 741 kicked off, as the chronicles of the time (at least, compared to the chronology of the genetic evidence) see it, by a fresh wave of settlement direct from Syria.⁸ So it seems to me that when we see a Middle Eastern genetic sample, it doesn’t have to be Jewish, and that ideally there would be some way to check this sample for what might be a tiny tiny Arabic representation, but might not (and it would be really nice to know which and where).

The bronze inscription of Botorrita, in eastern Ibero-Celtic characters

Then there is another missing population, which is the actual Iberians. Quite early on the authors decide that the best comparator for the African and Jewish samples is the Basque one, as it shows no or little mixture with those groups, so everything else in the peninsula is then thought of as being more or less of a mixture between the three `parental’ samples. Well, OK, but whenever the Basques arrived in Spain, other groups followed, most obviously the Celtic groups we now call Iberians, and also maybe some Visigoths, you know, though we don’t seem to credit that those were numerically significant any more.<sup>9</sup> I don’t think it diminishes the significance of the African and Jewish samples being different in the ways that they are too much, but I think that a better conceptual model might have been instead to take a total average of the peninsula and emphasise differences from it globally, rather than thinking in terms of `amibasqueornot’. Or, again, perhaps it would just be nice to have had some potential Celtic (or even Gothic) comparators factored in too so that we might get some sense of where those groups might have been best preserved, if they are at all. The paper’s only 11 pages long, after all, though I realise that I may just have idly asked for about three or four more years’ computing and sampling.

Interior of Santa María la Blanca, Toledo, previously a synagogue built in Almohad (i. e. a Berber Muslim) style. From Wikimedia Commons

So, in short, this stuff is really interesting but it’s very difficult to distill it down to historical events without essentially using what we already know to explain this new data. I get a certain kick out of knowing that some of the more traditional Reconquista-minded scholars would have been horrified to think that heroic Castile was actually more African and more Jewish than other areas of the peninsula but, if that’s down to post-reconquest resettlement by the kings, that becomes less of a delicious irony and more likely to be a reflection of the fact that populations who feel their identity may be dissipating are more likely to stress it aggressively. I think that these samples could actually be interrogated to tell us more about the settlement period by, for example, adding Arabic and Celtic comparators (if the latter can really be assembled, given how vague a group `Celtic’ populations are when considered historically¹⁰). At the moment, though, the main early medieval takeaway from this, which is what I at least am really interested in, is that it looks to be demonstrable that the African settler groups who (probably) arrived with and after the Muslim conquest really didn’t mix very much with the local populations. That’s not nothing, but I would still like to know if we might some day be able to guess at how much of the settling population they were from this kind of data, and thus guess also at the change in the élite too.

1. Susan M. Adams, Elena Bosch, Patricia L. Balaresque, Stéphane J. Ballereau, Andrew C. Lee, Eduardo Arroyo, Ana M. López-Parra, Mercedes Aler, Marina S. Gisbert Grifo, Maria Brion, Angel Carracedo, João Lavinha, Begoña Martínez-Jarreta, Lluis Quintana-Murci, Antònia Picornell, Misericordia Ramon, Karl Skorecki, Doron M. Behar, Francesc Calafell and Mark A. Jobling, “The Genetic Legacy of Religious Diversity and Intolerance: Paternal Lineages of Christians, Jews, and Muslims in the Iberian Peninsula” in The American Journal of Human Genetics Vol. 83 (Bethesda 2008), pp. 725-736, DOI 10.1016/j.ajhg.2008.11.007 (open access). Twenty authors seems like enough, really, although I can’t help feeling that they could also have credited the historian they consulted with (see below) and, after all, this is very far from the most extreme case of multiple authorship I can think of.

2. For the known history, they appealed to Dolors Bramon (ibid. p. 734, Acknowledgements), who is the current expert on what is to be learnt about Christian Iberian history from Islamic sources; her little anthology, De Quan Erem o No Musulmans: Textos del 713 al 1010. Continuació de l’Obra de J. M. Millàs i Vallicrosa (Vic 2000), is a frequent source of great help to me. So that was an unusually good choice, really, but apparently not a research contribution. Hmph.

3. Adams et al., “Genetic Legacy”, p. 727, citing E. Bosch, F. Calafell, D. Comas, P. J. Oefner, P. A. Underhill and J. Bertranpetit, “High-resolution analysis of human Y-chromosome variation shows a sharp discontinuity and limited gene flow between Northwestern Africa and the Iberian Peninsula” in American Journal of Human Genetics Vol. 68 (Bethesda 2001), pp. 1019–1029, and B. Arredi, E. S. Poloni, S. Paracchini, T. Zerjal, D. M. Fathallah, M. Makrelouf, V. L. Pascali, A. Novelletto and C. Tyler-Smith, “A predominantly neolithic origin for Y-chromosomal DNA variation in North Africa”, ibid. Vol. 75 (2004), pp. 338–345.

4. Adams et al., “Genetic Legacy”, p. 727 and describing work by Doron M. Behar and Karl Skorecki; the data is tabulated in the online version of the paper as “Haplogroups and Y-STR Haplotypes of Iberian Peninsula and Sephardic Jewish Samples” here (PDF).

5. Here as elsewhere, the sample from Asturias is just too small to allow significant conclusions, which may be just as well considering how much that one would expect is missing from it. This gives me pause, again, about drawing conclusions too far from the other areas with proportionally lower representation in the samples, including not least Minorca of course.

6. For the somewhat localised nature of the Andalusi state, I am used to citing Eduardo Manzano Moreno, La frontera de al-Andalus en la época de los Omeyas, Bibliotheca de Historia 9 (Madrid 1991), though I had the great pleasure of meeting the author this week and he tells me that he would now revise most of it! Extremely frustrating, as I have come to see it as canonical, which may be exactly why he would like to change it. Anyway. For place-names and settlement, the work of resort is by Pierre Guichard, either in French as Structures sociales « orientales » et « occidentales » dans l’Espagne musulmane (Paris 1977) or trans. into Castilian & rev. as Al-Andalus. Estructura antropológica de una sociedad islámica en Occidente (Granada 1998), though Dr Manzano tells me this too must be considered obsolete now. I don’t know if I’d agree there (or, it turns out, with quite a lot else Dr Manzano would argue, which was fun; more on this in due course). Compare Jessica A. Coope, “Marriage, Kinship, and Islamic Law in Al-Andalus: Reflections on Pierre Guichard’s Al-Ándalus” in al-Masaq Vol. 20 (London 2008), pp. 161-177, which is interesting because it disagrees with Guichard in exactly the opposite direction to Dr Manzano. For an English introduction to these issues, albeit a controversial one, see Roger Collins, The Arab Conquest of Spain, 710-797, History of Spain 4 (Oxford 1989). Collins caught it from the critics here because he effectively refuses to use Arabic historical writing, reckoning it all far too late and legendary to be anything other than misleading. There is also, of course, the fact that he doesn’t read Arabic, and this makes him an easy critical target because of course how can he know what he’s missing? but if you compare the exactly contemporary and much more traditional ‘Abdul Wahid Dhanun Taha, The Muslim Conquest and Settlement of North Africa and Spain (London 1989), I would say that it is fairly clear that Collins had a point. The fact that we can get three books like Collins, Taha and Guichard all purportedly telling the same story and disagreeing so incredibly (to say nothing of Manzano’s Frontera) is a measure of how charged these debates are. Without that charge, after all, how could we ever have had the now-legendary Ignacio Olagué, Les Arabes n’ont jamais envahi l’Espagne (Paris 1960, 2nd edn. 1973), to which cf. Pierre Guichard, “Les Arabes ont bien envahi l’Espagne : les structures sociales de l’Espagne musulmane” in Annales : Économies, sociétés, civilisations Vol. 29 (Paris 1974), pp. 1483-1513. I may have become sidetracked here.

7. Adams et al., “Genetic Legacy”, pp. 732-733.

8. Testified to even in the Christian Chronicle of 754, also known as The Mozarabic Chronicle though `Mozarab’ is one of those words that means too many things and should be retired, as translated in Kenneth Baxter Wolf, Conquerors and Chroniclers of Early Medieval Spain, Translated Texts for Historians 9 (Liverpool 1990, 2nd edn. 1999), pp. 111-160 with commentary pp. 25-42, cc. 82-86 in which the chronicler helpfully tells us that he wrote a whole book about this already so won’t repeat himself here. Do we have the book? No, we do not. Ah well. Nonetheless, it is this proximity to events that caused Collins to favour Christian sources over the Arabic ones. On the difficulties with the term `Mozarab’, see Richard Hitchcoock, Mozarabs in Medieval and Early Modern Spain: identities and influences (Aldershot 2008), passim but esp. pp ix-xx.

9. Roger Collins, Visigothic Spain 409-711, History of Spain 3 (Oxford 2004), pp. 25-26.

10. On which you can see the brief and bracing statements of Guy Halsall, who risks Godwin’s Law at an early stage of a book by comparing Celticism to Germanism in his Barbarian Migrations and the Roman West, 376-568, Cambridge Medieval Textbooks (Cambridge 2007), pp. 24-25.