“What connects archaeology and statistical physics?”, we asked ourselves one evening in The Marquis Cornwallis, a local Bloomsbury pub in London back in 2014, while catching up after more than a decade since our paths crossed last time. While bringing back the memories of that time we first met when we were both 16, it hit us that our enthusiasm for research we did as teenagers had not faded away: Jelena has always been occupied with game theory and networks, while Miljana has never stopped pursuing evidence for the world’s earliest metallurgy in the Balkans. Apart from sharing excitement that things at the end did not turn out too bad for us, two postdocs at the time at UCL and Imperial College London, we didn’t know details about each other’s research. Nevertheless, that winter night in The Marquis Cornwallis we decided to investigate the connection – not only for the sake of science, but for the thrills of reliving the days we spent together as teenagers in the famous Balkan ‘nerd hub’, Serbian Petnica Research Centre.
The first run of metals data through complex networks algorithms happened the night before we were due to deliver our ideas on how networks research can benefit early Balkan metallurgy research at a physics conference. Needless to say, we (over)committed ourselves to delivering a fresh view on this topic without giving it a thorough thought, rather, we hoped that our enthusiasm would do the job. That night, the best and the worst thing happened: our results presented the separation of modules, or most densely connected structures in our networks, as statistically, archaeologically, and spatiotemporally significant. The bad news was that we stumbled upon it in a classic serendipitous manner – we did not know what was it that we pursued, but it looked too good to let go. It subsequently took us three years to get to the bottom of networks analysis we made that night.
In simple terms, what we did here is present ancient societies as a network. A large number of systems can be represented as a network. For example, human society is a network where the nodes are people and the links are social or genetic ties between them. A lot of real-world networks exhibit nontrivial properties that we do not observe in a regular lattice or in the network where we connect the nodes randomly. For example social networks have the property called ‘six degrees of separation’, which means that the distance between any of us to anybody else on the planet is less than six steps of friendships. So any of us knows somebody, who knows somebody etc (six times) who knows Barack Obama or fishermen on a small island in Indonesia. Another property that is common in complex networks is so-called modularity. This means that some parts of the network are more densely connected with each other than with other parts of the network. Successful investigation of modularity or community structure property of networks includes detecting modules in citation networks, or pollination systems – in our case we used this property to shed light on the connections between prehistoric societies that traded copper. It turned out that they did not do it randomly, but within their own network of dense social ties, which are remarkably consistent with the distribution of known archaeological phenomena at the time (c. 6200- 3200 BC), or cultures.
What we managed to capture were properties of highly interconnected systems based on copper supply networks that also reflected organisation of social and economic ties between c. 6200 and c. 3200 BC.
Our example is the first 3,000 years of development of metallurgy in the Balkans. The case study includes more than 400 prehistoric copper artefacts: copper ores, beads made of green copper ores, production debris like slags, and a variety of copper metal artefacts, from trinkets to massive hammer axes weighing around 1kg each. Although our database was filled with detailed archaeological, spatial, and temporal information about each of 400+ artefacts used to design and conduct networks analyses, we only employed chemical analysis, which is the information acquired independently, and can be replicated. Importantly, we operated under the premise that networks of copper supply can reveal information relevant for the specific histories of people behind these movements, and hence reflect human behaviour.
Our initial aim was to see how supply networks of copper artefacts were organised in the past, and as the last step of analysis we planned to utilize geographical location only to facilitate visual representation of our results. Basically, if two artefacts from the same chemical cluster were found in two different sites, we placed a link between them. In the final step, the so-called Louvain algorithm was applied in order to identify structures in our networks, and we used it as a good modularity optimization method. Another advantage is this approach is that we can test its statistical significance and put a probability figure to the obtained modules.
What we managed to capture were properties of highly interconnected systems based on copper supply networks that also reflected organisation of social and economic ties between c. 6200 and c. 3200 BC. The intensity of algorithmically calculated social interaction revealed three main groups of communities (or modules) that are archaeologically, spatiotemporally, and statistically significant across the studied period (and represented in different colours in Figure 1). These communities display substantial correlation with at least three dominant archaeological cultures that represented main economic and social cores of copper industries in the Balkans during these 3,000 years (Figure 2). Basically, such correlation shows that known archaeological phenomena can be mathematically evaluated using modularity approach.
Although serendipity marked the beginnings of our research, our plan is to take it from here with a detailed research strategy plan, which now includes looking at other aspects of material culture (not only metals), testing the model on datasets across prehistoric Europe, or indeed different chronological periods. We can say that the Balkan example worked out well because metal supply and circulation played a great role in the lives of societies within an observed period, but it may not apply in cases where this economy was not as developed. The most exciting part for us though was changing our perspective on what archaeological culture might represent. Traditional systematics is commonly looking at cultures as depositions of similar associations of materials, dwelling and subsistence forms across distinct space-time, and debates come down to either grouping or splitting distinctive archaeological cultures based on expressions of similarity and reproduction across the defined time and space. But now we have the opportunity to change this perspective and look at the strength of links between similar material culture, rather than their accumulation patterns. This is a game changer for us. And we hope that this research inspires colleagues to pursue this idea of measuring connectedness amongst past societies in order to shed more light on how people in the past cooperated, and why.
Featured image credit: Mountains in Bulgaria by Alex Dimitrov. CC BY-SA 4.0 via Wikimedia Commons.