Featured post

Gender in the Merovingian World

Thursday, 15 December 2022

Archaeology, History and Bad Science: A critique of the analysis of DNA at Szólád (Hungary) and Collegno (Italy). Part 1 (Introduction; setting up the experiment).

 

[This is a critique I wrote to form my part of an article I co-wrote with Prof Martial Staub (Sheffield) about genetics and archaeology. Later I cut it down, with the idea of publishing the more detailed elements as 'online supplementary materials'. It didn't get published, sadly. I think that there were two main problems: one was a significant jump in the scale and breadth of Martial's part and mine; the other was that detailed critique like this is maybe what needs to happen before an article like that under discussion comes out, rather than afterwards. Nonetheless, I think this detailed critique is worth making, and if anyone does know of a journal or other publication that might be interested in considering this (or a shorter version of it), do please let me know.

I have split the piece into three shorter blogs for ease of reading. This part contains the introduction and discussion of the problems involved in the setting up of the 'experiment'; part 2 discusses methods and results; and Part 3 contains the overall conclusions, and an afterword.]

Introduction

As a more specific focus for our critique, we consider one particular study which made something of a media splash in 2018.[1] This thought-provoking piece compares evidence from a cemetery at Szólád near Lake Balaton in modern Hungary (Late Antique Pannonia) and another at Collegno in Piemonte, Italy, two areas linked historically by the sixth-century Longobard (or Lombard) Migration to Italy. The analyses revealed discrete groups within both cemeteries, showing different genetic traits, which were related to northern and southern Europe. These groups correlated with different methods of burial and (to a lesser degree) in terms of their diet.

When it appeared, the link to the online publication was retweeted by one Twitter account with the claim that ‘Peter Heather was right. The Völkerwanderung was a thing!’[2] We can leave aside for now the point that Heather certainly was right, for the very good reason that no serious scholar has ever denied that the ‘Great Migrations’ happened.[3] What requires more rigorous examination is whether this article supports the interpretation of the Barbarian Migrations that lies behind the term ‘Völkerwanderung’: not simply a large-scale, short-term population movement, but one that introduced novel cultural practices into the regions where the migrants eventually settled. More specifically, we can ask whether it does anything to help us ‘understand’ ‘6th-century barbarian social organization and migration’, as its title proclaims. Here the problem might be a disciplinary divergence in the meaning of the phrase ‘understanding society’ but, if viewed in a purely historical or archaeological sense, my answer will be definitive: no, it doesn’t. It describes two situations where people had (in some way) moved into a specific locality from elsewhere and had (possibly) used different burial practices. That is valuable. If, however, we want to understand those specific local phenomena and, a fortiori, if we wish to draw conclusions from them about the wider events of the ‘Barbarian Migrations’, I will demonstrate that we are no further on than when we were at the start. The only way in which the micro (local) and macro (pan-European) phenomena can be linked is via a number of uncritical assumptions, which the publication in question does nothing to interrogate. Those assumptions are, first, that if migration occurred then that must have been ‘barbarian’ and, second, that social organisation and variation in material culture must be explained by migration. In other words, the fact of ‘barbarian migration’ (which no one is denying) is the necessary and sufficient cause for late antique population and cultural change. It is in the current climate of ‘push-back’ against, and indeed flagrant misrepresentation of the views of,[4] those who deny a unique causal primacy to the Völkerwanderung, the enthusiastic expostulation about how science had proven that the Great Migrations were ‘a thing’ finds its context.

A more critical look at the analysis and its approaches is, however, necessary.[5] The popular reception of such studies and the nature of their media coverage are heavily grounded in suppositions that the methods of laboratory science are superior to the inevitably subjective assumptions and opinions of the historical disciplines and the partial – in both senses of the word – nature of their evidence.[6] Popularly, the epistemological claims of ‘hard science’ trump those of the arts and humanities. The Online Supplementary Materials (hereafter OSM) of the article under discussion set out the procedures employed in the laboratory analyses and mathematical modelling. I can see no reason to comment critically on these. It is no part of this discussion to cast – or even to imply – any doubts upon their quality and rigour. Whatever else one might say, it is an impressive piece of mathematical modelling and complex data analysis. My critique departs instead from consideration of the study’s rigour as a scientific experiment and thus confronts the broader epistemological claims. To anticipate my conclusions, my argument will be that – in that more general sense – the experiment represents bad science.

Setting up the experiment

The authors describe the experiment and its results thus:

[W]e obtained ancient genomic DNA from 63 samples from two cemeteries (from Hungary and Northern Italy) that have been previously associated with the Longobards, a barbarian people that ruled large parts of Italy for over 200 years after invading from Pannonia in 568 CE. Our dense cemetery-based sampling revealed that each cemetery was primarily organized around one large pedigree, suggesting that biological relationships played an important role in these early medieval societies. Moreover, we identified genetic structure in each cemetery involving at least two groups with different ancestry that were very distinct in terms of their funerary customs. Finally, our data are consistent with the proposed long-distance migration from Pannonia to Northern Italy. [Emphasis added]

The experiment was designed to confront the following questions:

Were specific barbarian peoples described in texts culturally and ethnically homogeneous populations, or were they ad-hoc and opportunistic confederations of diverse, loosely connected groups? What role did biological relatedness, being that of close kinship relations or long-term shared ancestry, play in the organization of these barbarian communities and how are such relationships related to patterns of material culture? Did this period involve long-distance migrations as described by late antique authors?

The most troubling aspect of the experiment concerns the way it was shaped by assumptions about its likely results. The OSM include a brief history of the Longobards but not one of late antique northern Italy and its long history of social, cultural, economic, and political relationships with transalpine areas. A reader of the article and its supplementary materials can be forgiven for believing – entirely incorrectly – that the Longobard movement was the only significant migration into Italy to occur in the period. Furthermore, the cemetery-sites investigated are located in the regions linked by historical accounts of the Longobard migration. In other words, the selection of data was made on a priori grounds. Now, genetic scientists might be led to believe that the epistemological status of the ‘Longobard’ label or identifier attached to the sites from which their samples were drawn was much more secure than it is in actuality, or that such a label might imply a genetically discrete population. They can understandably not be familiar with the broader historical and archaeological issues to which we shall return. One may therefore assume a ‘good faith’ procedure on their part in accepting certain parameters in the setting up of the experiment.

Fundamentally, though, one must ask where the control sample is. Assessment of the results’ significance requires comparable analysis of other sites either from the same period but not connected by the Longobard migration, or in the same regions but from a slightly earlier period. The main comparative aDNA evidence used was a pan-European range of samples from the Bronze Age, 1500 years or more before the Longobards’ migration.[7] This is insufficient. One needs comparanda from, say, the middle Roman period. A few early medieval samples were considered but their selection was problematic. A cluster taken from a 4th-7th-century English context was listed as ‘Anglo-Saxon’. Others, from the Caucasus, were labelled as being ‘Alans’. This raises questions in the context of a study of Barbarian Migrations as these names are late antique or early medieval ethnonyms not necessarily linked to genetics or to material culture in any straightforward way. In particular, the term ‘Anglo-Saxon’ implies that these subjects had migrated into the Empire, in this case into the diocese of Britanniae. When linked to the burials of supposed migrants in Hungary and Italy it risks giving the impression of a shared ‘barbarian’ origin.

Only six Italian aDNA samples were consulted in the whole analysis, all from prehistory, and three from the same site. On the basis of this lacuna alone, one is entitled to doubt the results’ validity. The authors state that what they consider to be ‘northern’ DNA is otherwise unknown in Italian samples.[8] This is illogical unless one assumes, a priori, that those graves with such DNA at Collegno were not of Italians. The only Italian aDNA from the relevant period to be considered was that at Collegno, so it is equally true to say that ‘northern’ DNA was found in every early medieval Italian population examined in the study. Without appropriate comparanda it is not possible to know how atypical this profile would be for a late antique cemetery in the Po valley. The only mitigation again concerns the information passed to the genetic scientists. If the assumption, made in good faith, was that Szólád and Collegno were both sites that could reasonably be accepted as having had Longobard populations, then the issue of a control might possibly have been seen as moot. If the Longobard nature of these cemeteries, their populations, and some of their material culture is taken as a given, its correlation with other variables being the subject of analysis, it does not need testing.

But therein lies the problem. On what basis is either Szólád or Collegno ‘Longobard’? The assignment of ethnic identities to forms of archaeological evidence has been hotly debated for over thirty years.[9] The Szólád/Collegno analyses were initially intended to contribute critically to this debate. While many archaeologists now accept that giving ethnic labels to forms of material culture is dubious, these reservations and the newer interpretations which they have engendered have been bitterly resisted; Longobard archaeology is one of the more conservative areas in this respect.[10] This problem even affects the sites’ dating. The earliest period of ‘Longobard’ archaeology begins in 568 precisely, because (so runs the argument) there were no Longobards in Italy before 568 and the material is Longobard. The ethnicity of the associated people is thus entirely prejudged. The same is true in Hungarian archaeology where, conversely, Longobard material cannot be later than 568. There is no archaeological reason to suppose that the first phase of ‘Longobard’ burials at Collegno could not belong to a period perhaps ten (or more) years before 568. Archaeological periodisation cannot be fine-tuned to that level of accuracy and the development of its techniques over the past forty years has shown that phases of change can drift back or forward by decades. For example, a key archaeological transition in artefact-forms which, in the 1980s, was dated to ‘c.600’ is now placed closer to 575/80.[11] The extent to which early Italian ‘Longobard’ material is still generally congruent with comparative material from the period after 568 is unclear. Without the influence of the historical record, some of it could unproblematically be placed earlier. Equally, there is no archaeological reason why Szólád could not have been used for a decade or more after 568. Overall, there are no prima facie archaeological grounds for supposing that the earliest phase at Collegno is not contemporary with the last period at Szólád, whether before, after, or straddling 568. This must influence how we employ cemetery evidence to think about trans-Alpine connections and the Longobard migration.

The link between this material and the Longobards’ arrival is, however, too well entrenched in most Italian archaeology to permit detailed or rigorous scrutiny of the idea. Indeed, the archaeology of this period is currently experiencing a backlash against attempts to reassess the role of Barbarian migration in material cultural change, further making critical revaluation unlikely. The lead-author of the OSM Collegno discussion has written a forceful – if unconvincing – defence of the traditional idea that ethnicity is manifest in late antique burial customs[12] and the lead-author of the Szólád discussion subscribes to a similar viewpoint.[13] The team involved in the Szólád-Collegno analyses included only archaeologists who subscribe to traditional ‘ethnic’ readings of material culture so no serious consideration was given to alternative interpretations.[14] Against that background it is extremely difficult to avoid the implication that incomers distinguished from locals are to be identified as Longobards.

We must also scrutinise the assumptions about the nature of population movement, which can be illustrated with three quotations:

While previous sampling from the era has been limited, we note that published fourth- to seventh-century genomes from Britain, Bavaria, Lithuania, and the Caucasus, analysed alongside our own ancient samples, cluster close to their modern counterparts.

We found no evidence that such ancestry was present in northern Italy during this time (who instead resemble modern southern and Iberian Europeans), which would be consistent with inferred long term barriers to gene flow in Europe across the Alps.

Modern European genetic variation is generally highly structured by geography.[15]

The fundamental assumption is of long-term population immobility, against which seemingly rare migration can be set and, one assumes, be clearly identifiable. Whatever the scientific grounds, historically it is a counterfactual. As will be discussed later, movement of individuals or large groups across the Alps, in both directions, has been constant since prehistory. Instead, support was drawn from an article claiming that the study of a person’s DNA allowed their geographical origins to be reliably estimated even to village level.[16] This conclusion was reached from analysis of populations from three islands off Scotland, three villages in the Alps, and three villages in Croatia. When separated by up to thirty miles of sea or several thousand metres of mountain, it is scarcely surprising that marriage tended to take place within each individual locality. The Croatian case study, where such natural barriers were absent, was far less conclusive. Another support was a study of the British Isles which claimed that modern British DNA showed the persistence of the political units of the ‘Dark Ages’[17] but actually came, very clearly, to the hardly staggering conclusion that people in low-lying arable areas find it easier to move around to find sexual partners than those living in the middle of mountain ranges.[18] Some of these issues resurface when considering the experiment’s methods. The logical accompaniment to the idea that difficult physical geography creates barriers to gene-flow – that areas of easier communication might accelerate it – seems not to feature anywhere in the cited literature. The methods of plotting the spatial distribution of particular genetic components[19] assumed a ‘flat earth model’ – in other words, a genetic feature would normally diminish across space in an even fashion. This contradicts the assumption that physical geography presents long term barriers to gene-flow, as indeed does the very wide geographical spread of modern subjects listed as Swiss.

Notes

[1] ‘Understanding 6th-century barbarian social organization and migration through paleogenomics’, in Nature Communications (2018) 9:3547 (DOI: 10.1038/s41467-018-06024-4). For interesting discussion of the media and aDNA studies, see, Källén, Anna & Mulcare, Charlotte & Nyblom, Andreas & Strand, Daniel. (2019). ‘Archaeogenetics in Popular Media: Contemporary Implications of Ancient DNA.’ Current Swedish Archaeology. 27. 69-91. 10.37718/CSA.2019.04. I am grateful to Oren Falk for this reference.

[2] This tweet or the account seems subsequently to have been deleted or has 'blocked' me.

[3] Susan Oosthuizen, The Emergence of the English, is the latest attempt to deny that there was an Anglo-Saxon migration but her argument is extremely problematic.

[4] G.P. Brogiolo, ‘Dati archeologici e beni fiscali nell’Italia Goto-Longobarda’ in Between Taxation and Rent: Fiscal Problems from Late Antiquity to the Early Middle Ages/Entre el Impuesto y la Renta. Problemas de la fiscalidad tardoantigua y altomedieval, ed. P.C. Díaz & I. Martín Viso (Bari, 2011), pp.87-105, at p.100; G.P. Brogiolo & A. Chavarría Arnau, ‘Chiese e insediamenti rurali tra V e VIII secolo prospettive della ricerca archeologica’ in ‘Ipsam Nolam barbari vastaverunt’: L’Italia e il Mediterraneo occidentale tra il V secolo e la metà del VI. Atti del Convegno internazionale di studi (Cimitile-Nola-Santa Maria Capua Vetere, 18-19 giugno 2009), ed. C. Ebanista & M. Rotili (Cimitile, Tavolario edizioni, 2010), pp.45-62, at 46-47.

[5] In 2019 I presented a preliminary critique of this piece at the University of Sheffield. The authors’ response, made before the lecture’s text was publicly available (at https://600transformer.blogspot.com/2019/03/the-barbarian-migrations-in-21st-century.html), was dismissive at best. Cp. the similarly aggressive response by the authors of Charlotte Hedenstierna-Jonson et al., “A Female Viking Warrior Confirmed by Genomics,” American Journal of Physical Anthropology 164.4 (2017): 853-60, to Judith Jesch’s similarly blogged response: “Let’s Debate Female Viking Warriors Yet Again,” Norse and Viking Ramblings (9 September 2017) http://norseandviking.blogspot.com/2017/09/lets-debate-female-viking-warriors-yet.html. Källén et al., ‘Archaeogenetics in Popular Media’, p.85. I am grateful to Oren Falk for letting me read his unpublished paper ‘Death and the Shield-Maiden: Viqueering Vikings and Viquens’, a detailed critical discussion of the supposed woman-warrior found at Bjirka and the ensuing debate.

[6] Källén et al., ‘Archaeogenetics in Popular Media’.

[7] Below, n.28.

[8] Amorim et al. ‘Understanding’, p.5.

[9] This debate was reignited by G. Halsall, ‘The origins of the Reihengräberzivilisation: Forty Years on.’, in Fifth-Century Gaul: A Crisis of Identity? Ed. J.F. Drinkwater and H. Elton (Cambridge, 1992), pp.196-207; G. Halsall, ‘Archaeology and the late Roman frontier in northern Gaul: the so-called Föderatengräber reconsidered’, in Grenze und Differenz im früheren Mittelalter, ed. W. Pohl & H. Reimitz (Vienna, 2000), pp.167-80. These studies are reprinted in G. Halsall, Cemeteries and Society in Merovingian Gaul. Selected Studies on History and Archaeology, 1992-2009 (Leiden, 2010), pp.93-130, alongside ‘Commentary 2: Careful with that axe, Eugenius’ (pp.131-67), which responds to critiques made up until about 2010. For similar criticisms of the traditional ethnic reading, see above all, S. Brather, Ethnische Interpretationen in der frühgeschichtliche Archäologie (Ergänzungsbande zum Reallexikon der germanischen Ältertumskunde 42: Berlin, 2004); P. von Rummel, Habitus Barbarus: Kleidung und Repräsentation spätantiker Eliten im 4. und 5. Jahrhundert Ergänzungsbande zum Reallexikon der germanischen Ältertumskunde 55: Berlin, 2007); F.J. Theuws & M. Alkemade, ‘A kind of mirror for men: sword depositions in late antique northern Gaul’, in Rituals of Power. From Late Antiquity to the Early Middle Ages, ed. F.J. Theuws & J.L. Nelson (Leiden, 2000), pp.401-76; F. Theuws, ‘Grave-goods, ethnicity and the rhetoric of burial sites in late antique northern Gaul’, in Ethnic Constructs in Antiquity. The Role of Power and Tradition, ed. T. Derks & N. Roymans (Amsterdam, 2009), pp.283-319.

[10] See the works by Brogiolo, above n.4; V. Bierbrauer, ‘’Zur ethnischen Interpretation in der frühgeschichtlichen Archäologie’, in Die Suche nach den Ursprüngen von der Bedeutung des frühen Mittelalters (Vienna, 2004); Kazanski, M., & P. Périn, 2008. ‘Identité ethnique en Gaule à l’époque des Grandes Migrations et des Royaumes barbares: étude de cas archéologiques’.  Antiquités Nationales 39, pp.181-216. the articles by Giostra and Vida cited below, n.12

[11] R. Legoux, P. Périn & F. Vallet, Chronologie Normalisée du Mobilier Funéraire Mérovingien entre Manche et Lorraine (3rd revised edition; Condé-sur-Noireau, 2009).

[12] C. Giostra ‘Goths and Lombards in Italy: the potential of archaeology with respect to ethnocultural identification’, Post-Classical Archaeologies 2011, pp.7-36. There is no space for detailed refutation of the argument here.

[13] T. Vida, ‘Conflict And Coexistence: The Local Population Of The Carpathian Basin Under Avar Rule (Sixth To Seventh Century)’, in The Other Europe in the Middle Ages. Avars, Bulgars, Khazars and Cumans, edited by F. Curta, Brill, Leiden, 2008, pp. 13–46. Vida dismisses Brather’s work: ‘Brather’s position may be popular with advocates of a post-processualist critique of both archaeological sources and the methods of the archaeological inquiry, but it rests on wrong assumptions’ (p.15). He presents no substantive reasons why this might be the case. He and Giostra both deploy ‘post-processualism’ – a very loose archaeological school of thought of which neither seems to have a clear understanding – as a sort of bogeyman, and both rely on the heavily criticised notion of a pan-Germanic culture. For recent discussion of the latter, see M. Friedrich & J. Harland (ed.) Interrogating the “Germanic”: A Category and its Use in Late Antiquity and the Early Middle Ages.

[14] The acknowledgements of Amorim et al., ‘Understanding’, include thanks, for ‘helpful’ conversations, to Dr Philipp von Rummel and Professor Frans Theuws, both of whom have published rigorous critiques of the traditionalist ‘ethnic’ reading of grave-goods (above, n.9). These conversations have clearly been ignored.

[15] Amorim et al. ‘Understanding’, p.5, p.5, p.9, respectively.

[16] O’Dushlaine, C., et al. ‘Genes predict village of origin in rural Europe.’ Eur. J. Hum. Genet. 18, 1269–1270 (2010).

[17] Leslie, S., et al. ‘The fine-scale genetic structure of the British population.’ Nature 519, 309–314 (2015).

[18] Readers should note here the significant difference between the proposition that mountain ranges act as barriers to gene-flow across them and the claim that mountains constrain gene-flow among the population living within the range.

[19] Yang, WY., Novembre, J., Eskin, E., & Halperin, E., A model-based approach for analysis of spatial structure in genetic data. Nat Genet 44, 725–731 (2012). https://doi.org/10.1038/ng.2285