Sometimes conferences can be quite predictable: I know who I will meet, I know what I will hear, I know where I will get a drink at the end of the day. The Tracing Networks conference held at the British Academy two weeks ago was not one of those predictable events, for a number of reasons. First of all, because I forgot all about it. I woke up one day and noticed two days of Tracing Networks in my calendar. I arrived at the venue without having a clue who would be there, who would present, what they would be talking about and where I could get a drink. And I can definitely recommend forgetting about conferences to everyone, because the event turned out to be a very enjoyable experience.
Lin Foxhall giving the opening address, with Anthony Harding
Lin Foxhall gave the opening address in name of the Tracing Networks
team. Her talk was an overview of the project, and their search for a suitable methodological framework. This self-reflective and honest discourse was really fascinating. Lin went through a range of arguments why actor-network theory and formal network methods were not suitable. She said that network perspectives are good to think with but meaningfully and rigorously applying them within an archaeological context is particularly difficult. In my opinion this is totally true and cannot be emphasized enough. The team found a method based on ontologies and semantic web most appropriate for dealing with the large and very diverse datasets the project is concerned with.
Another presentation that interested me was Borja Legarra Herrero’s talk on using SNA for studying social change in Late Bronze Age Southern Spain. Some of his slides and parts of his paper revealed a very useful side of networks: their ability to communicate simple but useful structural ideas as small graphs representing different extreme hypotheses (e.g. star graph vs line graph vs circle graph). The usefulness of networks as a tool for communication is often uncritically exaggerated. I learned from experience that showing people real networks representing real data results in awkward silences: people don’t get it. True, these graphs become extremely useful once you understand the layout algorithm and play around with alternative visualizations. But their ability to communicate simple ideas is trivial compared to simplifying graphs of just a few nodes and links.
Leif Isaksen at Tracing Networks
This issue came up again during Steve Conway’s
reflections on graph visualizations. His paper took my own article in Oxford Journal of Archaeology
as a starting point, and tried to find similar trends to the ones I described in his review of the use of formal network methods in the managerial literature. He identified some really familiar sounding issues: there is a tendency to conflate time; a tendency to ossify, to make static; an over-emphasis on the overall network and ignoring individual nodes; a tendency to let the network visualization speak for itself; and an under-emphasis on context. These are all common issues with the use of network visualizations, which are never neutral and are as laden with decisions and assumptions as any other communication medium (Steve wrote an interesting article about this in the british journal of management
). This does not mean network visualizations are useless, or even bad at what they do. One just needs to approach and use them with as informed an understanding as possible of the decisions and assumptions that went into their creation.
Another paper that interested me was delivered by Peter Van Dommelen. He opened his talk on a sobering note, stating that “networks are not everything, we need to understand what is going on inside the nodes themselves”. Peter was mainly concerned with developing a critical archaeological approach to the study of migrations, stressing that the context of migrations need to be understood. He argued that there was a reluctance to discuss migration in archaeology since two decades because earlier migration studies were overly simplistic. That’s why we need to look beyond and below networks, we need to contextualize migrations, because the arrows on a map approach is just not good enough. We don’t just want to trace the large-scale, possibly state-enforced networks, but also the personal small-scale networks. We need a focus on communities on the ground if we want to understand what is going on inside the nodes. It is in the end the people who matter, they did not just trace but created the networks we are talking about. Peter discussed his ideas in the context of Nurraghic culture in Sardinia. He is of course right, but I have the impression that up til now the people that are “doing networks” have tended to go for the big datasets evidencing large-scale patterns, because there is just such a good fit with the network methods. However, this means that the challenge of local-scale, more contextualized archaeological network analysis remained under-explored.
Carl Knappett at Tracing Networks
… Until now? Carl Knappett clearly did not shy away from more small-scale and contextualized network approaches. His paper provided a balanced overview of network methods and theories, of the issues involved and the potential gains of a networks perspective for archaeology. He argued that network analysis in archaeology works best if node selection is unproblematic. It imposes some sort of order over a messy dataset. Although this is undeniably the case, it has to be said that some archaeologists are making real progress in confronting this issue. Ethan Cochrane and Carl Lipo
explore how different artefact classifications emerge when different network approaches are used. In his PhD thesis Matt Peoples
compares networks of ceramics classified by traditional ware typologies with networks of ceramic technical features. Carl continues by stating the importance of node definition and that this is a theoretical decision, i.e. it is wrong to think that SNA is untheoretical (Carl referred to Butts’ 2009 paper in Science
). I could not agree more. The decisions an archaeological network analyst makes when defining nodes and edges, when selecting or modifying analytical techniques and when interpreting the outcomes are fundamentally influenced by their theoretical preconceptions. When I was recently tackling this issue for one of my case-studies I challenged myself to come up with at least two different ways of creating a network from the same dataset; in the end I found ten! Other issues raised by Carl involved temporal and geographical scales. He claimed that although archaeological network methods are often static, this is not a problem of the network perspective per se. In fact, the meaning of nodes or categories of analysis can emerge through the process of thinking through networks (Carl referred to Astrid Van Oyen’s
work on comparing ANT and SNA). Carl challenged many of these issues head-on through his case studies from Bronze Age Crete, which revealed exactly how challenging they really are.
The Tracing Networks conference was a great experience, not in the least because I was genuinely surprised to see so many scholars there with shared interests doing fascinating work. I am looking forward to the proceedings and to forgetting about some of the upcoming conferences in my calendar.