The networks they are a-changin’: introducing ERGM for visibility networks

July 17, 2014

legosIn my madness series of posts published a few months ago I mentioned I was looking for a method to study processes of emerging intervisibilty patterns. I can finally reveal this fancy new approach to you 🙂 Here it is: introducing exponential random graph modelling (ERGM) for visibility networks. In previous posts I showed that when archaeologists formulate assumptions about how lines of sight affected past human behaviour, these assumptions imply a sequence of events rather than a static state. Therefore, a method is needed that allows one to test these assumed processes. Just analysing the structure of static visibility networks is not enough, we need a method that can tackle changing networks. ERGM does the trick! I just published a paper in Journal of Archaeological Science with Simon Keay and Graeme Earl that sets out the archaeological use of the method in detail. You can download the full paper on ScienceDirect, my Academia page or via my bibliography page. But in this blog post I prefer to explain the method with LEGOs 🙂



Social network analysts often use an archaeological analogy to explain the concept of an ERGM (e.g., Lusher and Robins 2013, p. 18). Past material remains are like static snapshots of dynamic processes in the past. Archaeologists explore the structure of these material residues to understand past dynamic processes. Such snapshots made up of archaeological traces are like static fragmentary cross-sections of a social process taken at a given moment. If one were to observe multiple cross-sections in sequence, changes in the structure of these fragmentary snapshots would become clear. This is exactly what an ERGM aims to do: to explore hypothetical processes that could give rise to observed network structure through the dynamic emergence of small network fragments or subnetworks (called configurations). These configurations can be considered the building blocks of networks; indeed, LEGO blocks offer a good analogy for explaining ERGMs. To give an example, a network’s topology can be compared to a LEGO castle boxed set, where a list of particular building blocks can be used to re-assemble a castle. But a LEGO castle boxed set does not assemble itself through a random process. Instead, a step by step guide needs to be followed, detailing how each block should be placed on top of the other in what order. By doing this we make certain assumptions about building blocks and their relationship to each other. We assume that in order to achieve structural integrity in our LEGO castle, a certain configuration of blocks needs to appear, and in order to make it look like a castle other configurations will preferentially appear creating ramparts, turrets, etc. ERGMs are similar: they are models that represent our assumptions of how certain network configurations affect each other, of how the presence of some ties will bring about the creation or the demise of others. This is where the real strength of ERGMs lies: the formulation and testing of assumptions about what a connection between a pair of nodes means and how it affects the evolution of the network, explicitly addressing the dynamic nature of our archaeological assumptions.

More formally, exponential random graph models are a family of statistical models originally developed for social networks (Anderson et al. 1999; Wasserman and Pattison 1996) that aim to scrutinize the dependence assumptions underpinning hypotheses of network formation by comparing the frequency of particular configurations in observed networks with their frequency in stochastic models.

The figure below is a simplified representation of the creation process of an ERGM. (1a) an empirically observed network is considered; (1b) in a simulation we assume that every arc between every pair of nodes can be either present or absent; (2) dependence assumptions are formulated about how ties emerge relative to each other (e.g. the importance of inter-visibility for communication); (3) configurations or network building blocks are selected that best represent the dependence assumptions (e.g. reciprocity and 2-path); (4) different types of models are created (e.g. a model without dependence assumptions (Bernoulli random graph model) and one with the previously selected configurations) and the frequency of all configurations in the graphs simulated by these models is determined; (5) the number of configurations in the graphs simulated by the models are compared with those in the observed network and interpreted.


My madness series of posts and the recently published paper introduce a case study that illustrates this method. Iron Age sites in southern Spain are often located on hilltops, terraces or at the edges of plateaux, and at some of these sites there is evidence of defensive architecture. These combinations of features may indicate that settlement locations were purposefully selected for their defendable nature and the ability to visually control the surrounding landscape, or even for their inter-visibility with other urban settlements. Yet to state that these patterns might have been intentionally created, implies a sequential creation of lines of sight aimed at allowing for inter-visibility and visual control. An ERGM was created that simulates these hypotheses. The results suggest that the intentional establishment of a signalling network is unlikely, but that the purposeful creation of visually controlling settlements is better supported.

A more elaborate archaeological discussion of this case study will be published very soon in Journal of Archaeological Method and Theory, so stay tuned 🙂 Don’t hesitate to try out ERGMs for your own hypotheses, and get in touch if you are interested in this. I am really curious to see other archaeological applications of this method.

References mentioned:

Anderson, C. J., Wasserman, S., & Crouch, B. (1999). A p* primer: logit models for social networks. Social Networks, 21(1), 37–66. doi:10.1016/S0378-8733(98)00012-4

Lusher, D., Koskinen, J., & Robins, G. (2013). Exponential Random Graph Models for Social Networks. Cambridge: Cambridge university press.

Lusher, D., & Robins, G. (2013). Formation of social network structure. In D. Lusher, J. Koskinen, & G. Robins (Eds.), Exponential Random Graph Models for Social Networks (pp. 16–28). Cambridge: Cambridge University Press.

Wasserman, S., & Pattison, P. (1996). Logit models and logistic regressions for social networks: I. An introduction to Markov graphs and p*. Psychometrika, 61(3), 401–425.

Archaeological and historical network analysts unite!

July 10, 2014

315px-I_Need_You_on_the_Job_Every_Day_-_NARA_-_534704Network science is becoming more commonly applied in both archaeology and history. But this is not happening without difficulties. Pioneers in both disciplines are now trying to overcome the numerous challenges that still surround their use of network techniques: how to deal with fragmentary data, performing analyses over extremely long time spans, using material data in network science to understand past human behaviour, …. I believe archaeologists and historians should face these challenges together! Through collaboration we might come to a better understanding of the use of network science in our disciplines much faster. In a recently published article in Nouvelles de l’Archéologie, Anna Collar, Fiona Coward, Claire Lemercier and myself show how many of the challenges that archaeologists and historians have identified are actually not discipline-specific: we CAN collaborate to tackle them together. Since this article is in French I wanted to provide an English summary of our argumentation here (written with my co-authors). The full article can be downloaded on Academia or through my bibliography page.


One of the key aspects of historical sources, compared to archaeological sources, is that the former often allow for the identification of past individuals, by name, and by role. This richness of data at the individual level means that network analytical methods can be very powerful in the illumination of past social networks and the details of particular places and times – offering, where the data are good enough, a window onto past social lives and interactions, and allowing the synchronic analysis of social networks at a particular moment in time.

However, the issues most commonly mentioned by historical network analysts also concern problematic and incomplete data. These issues are undeniably more significant for archaeology and history than for contemporary social sciences such as sociology. But we should not overestimate their potential impact. Even sociological research in contemporary populations face similar issues where full data may not be available for a variety of reasons, and although the problems are clearly more fundamental in history and archaeology, this also means that researchers in both disciplines have long been accustomed to dealing with, and developing methods at least partially compensating for, partial and biased datasets. As a result, this may be one important area where archaeology and history can contribute its expertise to other disciplines working with imperfect network data.


In contrast to history, archaeology is much less frequently furnished with such focused evidence. In archaeology, individuals are typically identified indirectly through the material remains they leave behind, and even where they can be identified, they often remain without names or specified roles.  Not only is archaeological data typically not ‘individualized’, but it can also rarely be attributed an exact date. Most archaeological data typically has date ranges with differing probabilities attached to them, making the establishment of contemporaneity between entities/potential nodes in networks (e.g. individuals; events; settlements) highly problematic. Because of this, archaeologists have tended to focus on the synchronic study of human behavioural change over the long-term, rather than on the diachronic examination of behaviour and interaction. A further characteristic of archaeological data is that it is also likely to be more strongly geographically grounded. Indeed, the geographical location of archaeological data is often among the few pieces of information archaeologists possess. Finally, network analytical methods in archaeology tend to focus most closely on long-term changes in the everyday lives of past peoples.

Common challenges in archaeology and history

Alongside these differences, there are also a number of common challenges facing archaeology and history, as ultimately both disciplines aim to achieve similar goals relating to understanding past interactions and processes.

The most significant of these common challenges are the fragmentary datasets that often characterize both disciplines; we typically deal with bad samples drawn from populations of unknown size and/or with unknown boundaries, snapshots of the past that are heavily biased by differential preservation and/or observation effects. However we argue that this does not exclude the use network techniques in our disciplines, nor does it limit us to only those research contexts for which high quality datasets are available.

A second issue facing our disciplines is that many methodological and theoretical network approaches have been developed in other disciplines to address particular research themes. As a result, they therefore function according to certain rules and/or have certain specific data requirements that might prevent straightforward applications in our disciplines.

Furthermore,  using a network approach to study a past phenomenon necessarily requires a researcher to make a series of decisions about how the parameters of that phenomenon should be represented – for example, what entities to use as nodes and what forms of relationship to model as vertices. Archaeologists and historians familiar with the analytical and visualization techniques used by researchers studying modern phenomena may find many analytical approaches and visualization techniques that are not appropriate or achievable. The past phenomena we are interested in, the kinds of questions our data allows us to ask, and the often very specific parameters of human behaviour assumed by archaeologists and historians for investigating the past, are likely to mean we will ultimately need to develop purpose-made visualization and analysis techniques. At the least we will need to acquire a critical understanding of the various methods available if we are to represent archaeological and historical network  data in appropriate ways – and indeed, to ‘read’ such visualizations and analysis results correctly.

Finally, the poor chronological control characteristic to a certain extent of historical and to a much greater extent of archaeological datasets, limits our knowledge regarding the order in which nodes and links in networks became salient and also the degree of contemporaneity between nodes. This is likely to have significant ramifications for the ways in which archaeologists and historians visualize and analyse networks, driving a need to consider ‘fuzzy’ networks, margins of error and probabilistic models, as well as the consideration of complex processes of network change and evolution over time.

Unite! Meeting the challenges together

In the recent surge of network applications in archaeology and history, it would seem that the two disciplines have thus far focused their efforts on the more obvious potential applications which mirror those most common in other disciplines, such as the identification and interpretation of ‘small-world’ network structure or the choice of datasets that are readily envisaged as or translated into network data (e.g. road and river networks). Such analyses have demonstrated the potential of the methods for archaeological and historical datasets; however, we believe that potential applications go far beyond this, and that network approaches hold a wealth of untapped potential for the study of the past. To achieve this potential, we will need to become more critical and more creative in our applications, and explore not simply what network science can offer the study of the past, but also what our disciplines offer in terms of developing that science – firstly to tackle specifically archaeological and historical questions, but ultimately to broaden the scope of the science itself as methodologies specifically developed for use in archaeological and historical contexts are taken up for use in tackling similar questions in other disciplines.

TCP (2013_05_12 19_17_14 UTC)Initiatives like The Connected Past and Historical Network Research offer a platform that would allow for exactly this kind of interaction between network scientists and those applying network science to the study of the past. The challenges individual members were encountering in our own research across archaeology and history encouraged us to consider developing a mutually supportive space in which to share concerns and problems, and to discuss ideas and approaches for moving beyond these.

We suggest that simply bringing people together through conferences, workshops, conference sessions and more informal groupings is key to fostering the dialogue between the disciplines that is so important to move forward applications of network analysis to the study of the past. Talking to each other across traditional disciplinary boundaries is vital in the ongoing development of network perspectives on the past. However, as noted above, at the same time we also need to be more sensitive to the specific demands of our disciplinary goals and our datasets and develop new network methods that suit our disciplines better. The sociological roots of most social network analysis software packages means that these are often designed and engineered to address discipline-specific research concepts that may not be appropriate for archaeology and history. SNA software has generally been created to deal with interactions between people in a modern setting – where the individual answers to questions about interactions can be documented with a degree of accuracy. As such, this software and network methodologies in general will need to be applied with care and ideally even developed from scratch for use with networks comprised of nodes which are words, texts, places or artefacts, for the characteristically fragmentary and poorly chronologically controlled datasets of archaeology and history, and for research that aims to go beyond the structuring of individual networks of contemporary nodes to investigate questions of network evolution and change. While interdisciplinary dialogue is crucial, we will need to be sensitive to the discipline-specific idiosyncracies of our data and to critique rather than adopt wholesale practices used in other fields. In this way, rather than apologizing for the ‘deficiencies’ of our datasets in comparison with those characteristic of other disciplines, we will also be able to make novel contributions to the wider field based on the new questions and challenges the study of the past offers network science.

Two months of insanity

March 10, 2014

Lindroth_The_Absent-minded_ProfessorIt’s finally there: the last two months of my PhD. Ever since I started almost four years ago everyone I talked to with a Dr. in front of their name told me the same thing, that the last few months are the hardest. It sounded as if when you finally decided to finish the damn thing off it starts putting up a fight. This usually finishes in the valiant PhD student winning the battle but loosing part of their sanity and most of their short-term memory in the process. My short-term memory is long gone (this is the main reasons why I claim to show promise for a career as an absentminded academic), but I have held on to my sanity. So far.

As I am working my way through my PhD in the coming two months I will document my struggle and loss of sanity on this blog, hoping it will end in victory. You can expect blog posts about all of the case studies I worked on in the last few years. In particular citation networks and visibility networks. But I will also share some of the conclusions I drew from working with network methods as an archaeologists, the challenges archaeologists are faced with, how we could confront these challenges, and my efforts to make a small contribution towards this. So stay tuned, and above all, please don’t hesitate to comment and provide me with your feedback on my work. I can use it now more than ever! 🙂

On organising the Hestia2 seminar in Southampton

July 29, 2013
John Goodwin presenting at Hestia2 in Southampton

John Goodwin presenting at Hestia2 in Southampton

As the organiser of the Hestia2 seminar in Southampton I could write about our initial struggle to find a good format, my fight with the university to book a seminar room in a completely booked out campus, discussions with our financial support staff to figure out a balanced budget, the technical flaws with our livestream feed, and of course the many very human feelings like “no-one will turn up!?!” and “what shirt should I wear?”. But none of that would be very interesting to read, and all of these concerns are now firmly pushed to the back of my mind and replaced by the feeling that this seminar was a success!

It will not come as a surprise that the organiser thinks his own event was a success. So let me at least try to come up with some objective-sounding arguments why this was in fact the case.

Multi-disciplinarity: organising a multi-disciplinary event is always risky. You need to address a very diverse target audience and convince them that the topics covered and the discussions will be of interest. People with different backgrounds also tend to talk in different languages: physicists will talk “maths”, classicists will talk “Greek/Roman”, archaeologists will talk “stuff”. The Hestia2 seminar was such a multi-disciplinary event. It was attended by classicists, historians, archaeologists, physicists and designers from the academic, commercial and governmental sectors. Despite this diversity the discussions very often converged into common interests. These included how large datasets like those held by the Ordnance Survey and the Historic Environment Records (HERs) could be usefully combined using new technologies, or how uncertainty about data can be formally expressed and visualised. Finding such common grounds was very much thanks to the chairs of our session. For example, Max Schich confronted the multi-disciplinary background of our audience directly when he asked to what extent individuals need to have skills and knowledge traditionally associated with different disciplines and professions to allow them to apply linked data and network techniques critically and usefully. This question drew very diverse reactions from the audience. Some felt that our educational system should allow for complete diversity and customisations of skills and knowledge, others (and I am part of this particular camp) believe that collaboration is the key, that field specialists should remain specialists but be able to collaborate with specialists in other fields by having some very basic understanding of the other’s “language”, approaches and questions.

Exploration: Hestia2 is not about showing off a great piece of work our team did a few years ago. It’s about learning from different projects’, institutions’ and individuals’ experiences with using innovative technologies to understanding conceptions of space. It’s about exploring the potential of such techniques for providing innovative insights into old datasets, or for allowing us to ask new questions of our data. The Southampton seminar definitely had that exploratory vibe. Very different techniques and projects were presented. The first three talks very much set the scene by giving an overview of different approaches. Max Schich introduced us to networks, Alex Godden provided an insight into the issues surrounding the aggregation and management of historical/archaeological data, and John Goodwin showed how the Ordnance Survey (OS) is implementing linked data. The discussions that followed showed a genuine interest in innovative approaches but also a constant concern with getting at the fundamental issues that keep all this innovation together. For example, in our discussion we never restricted ourselves to asking how something could be done, but always focused on why we should do it in the first place. The question of why HER data could not be seamlessly linked with OS data, for example, was not because of technological restrictions but concerns about protecting cultural heritage and also commercial concerns. Once such concerns were addressed we turned our attention to how combining such diverse datasets could allow us to ask new research questions, or could lead to a better management of historical resources.

Weather: it has sunny and hot. That makes every event an instant success!

I am very much looking forward to the second Hestia2 seminar in Stanford, where I will be able to put my feet up a little and enjoy another round of stimulating multi-disciplinary exploration.

Connected Past workshop registration open

June 25, 2013

TCPWorkshop Announcement:


Networks offer one of the newest and most exciting approaches to archaeological and historical data analysis, and over the last two years, the The Connected Past team has brought together scholars from across the globe to discuss their research, with a session at Birmingham TAG 2011, the Southampton conference in March 2012, a session at the SAAs in Hawaii in April this year, and a collaboration with HESTIA this coming July.

But we’re also aware that starting to do network analysis isn’t always easy. It can be difficult to know which software to use, how to present data, what questions to ask, and what results really show. Because it’s hard for researchers at all levels who are starting to think about network analysis, we are delighted to announce that we have put together a programme for a two-day practical workshop at the University of Southampton on 17-18 September 2013.

The cost of the workshop is £20. PLACES ARE LIMITED TO 20. To register your interest, please email with a short statement detailing why you want to participate. We will be in touch once the registration deadline (22nd July) has passed. The programme can be found below and on The Connected Past website.

In addition, for those who want to overdose on networks, Southampton will also be hosting the 12th Mathematics of Networks meeting on 16th September. It’s very multi-disciplinary, with a focus on social science applications and the technical side of things.


Tuesday 17th September

• Introduction to networks in archaeology and history
• Preparing data for network analysis
• network creation and visualisation
• Archaeological and historical case studies
• Round table discussion
Reception at the Institute for Complex Systems Simulation

Wednesday 18th September

• Network analysis software
• Analysing network structure
• What method to use?
• Geographical network techniques
• Issues in archaeological and historical network analysis

• Andy Bevan (UCL)
• Tom Brughmans (Southampton)
• Anna Collar (McDonald Institute, Cambridge)
• Fiona Coward (Bournemouth)
• Marten Düring (Nijmegen)
• Claire Lemercier (Sciences-Po, Paris)
• Angus Mol (Leiden)

Digital Classicist video online

June 19, 2013

tomDCThe Digital Classicist people edited the video of the talk I gave a week ago (they do work fast). It’s available on their blog and on my bibliography page. Enjoy 🙂

Introducing ‘A Connected Island?’: how the Iron Curtain affected Archaeologists

September 20, 2012

Eötvös Loránd University (University of Budapest)

Eötvös Loránd University (University of Budapest)

After the Second World War the Iron Curtain sliced through the very centre of Europe forming a very real divide in both political and daily lives. In the second half of the 20th century the Soviet regime introduced a new structure to the academic institutions to countries like Poland, Hungary and former Czechoslovakia, including restrictions on contacts with the Western world and ideological pressure previously unknown in these parts of Europe. How did this situation affect researchers on both sides? Was Central European Academia really isolated from western influences? A new project funded by SotonDH aims to address this issue using Palaeolithic archaeology as a case study. In ‘A connected island? Evaluating influence and isolation of Central European Palaeolithic researchers during communism’ ACRG members Iza Romanowska and Tom Brughmans combine a traditional historiography with novel citation network analysis techniques to approach this issue from a new angle.

Isolated or not?

A heated debate has been taking place in Central European archaeology in the last two decades regarding the issue of isolation (or the lack thereof) from western influences during the second half of the 20th century. Difficulties related to obtaining the necessary passports and visas, the disparity in the values of currencies, and only limited formal international links between research institutions restricted research visits, data collection, literature review, and conference attendance. Equally hindering was the limited circulation of Western archaeological journals within the Soviet Bloc countries, and restricted accessibility to archaeological publications in general. This could have been further aggravated by language barriers and, to some extent, different disciplinary interests. All this does not necessarily mean that Central European researchers were completely unaware of what was happening in the West, as if living on an island unconnected to the rest of the world and immune to external influences.

It is difficult to quantitatively determine to what degree these limitations affected Central European researchers. The project team argues that citation data might allow (at least in part) for such a quantitative evaluation. When a researcher cites the work of another scholar they express in a very formalised way that they were influenced by this person. Citations are like handy proxies for tracing lines of knowledge dissemination and academic influence, obviously not fully representative for these very complex processes, but well suited to quantify the ‘awareness’ of other peoples research.

‘A connected island?’ will collect and explore citation data for Central European Stone Age studies, a relatively small but highly international research field that forms a well-defined case-study suitable for quantitative analysis. The project will initially focus on the Lower and Middle Palaeolithic of Poland, former Czechoslovakia and Hungary. The citation behaviour of scholars working in these countries will be confronted with that of Western European Palaeolithic researchers. The proposed project therefore aims to explore the degree of interaction and academic influence between Central and Western European researchers in Lower and Middle Palaeolithic archaeology during communism (1945-1989) through citation network analysis, in order to evaluate the hypothesis that Central European researchers worked in strong academic isolation.

Data collection in Hungary

Digital citation datasets are available online through services like Google Scholar or Web of Knowledge. However, neither of these is very comprehensive for books or local non-peer-reviewed journals where a lot of the Palaeolithic archaeology of Central Europe was published in. So despite of the revolution in digital data collection brought about by the World Wide Web, a critical analysis could not be performed without visiting key libraries and research institutions in Central Europe. So the ‘connected island’ team hit the road.

The first phase of data collection was conducted in May 2012. In this first round a visit was paid to the Institute of Archaeology at the Jagiellonian University in Krakow. Thanks to a bursary from SotonDH the team could perform the second phase of data collection at the Hungarian Academy of Sciences and the Institute of Archaeological Science at the Eötvos Loránd University, both in Budapest. The third phase will aim to collect relevant literature from the Ústav pro pravěk a ranou dobu dějinnou at Charles University in Prague.

A few preliminary results

The widely held assumption that archaeological data from Central Europe was published in local languages is false. At least half, if not more, of Central European archaeology publications from this period were published in either German, French or English alongside the national language. The image that all countries under the influence of the former Soviet Union published in Russian is incorrect.

Just like their Western European or American colleagues Central European Palaeolithic researchers worked within a very strong ‘Bordean’ framework (named after a famous French researcher: Francois Bordes and his wife Denisse de Sonneville-Bordes).

From a cursory check, one gets an impression that Palaeolithic researchers in Central Europe were well informed of the developments on the other side of the Iron Curtain and quoted western authors extensively. The same can be said in reverse for their Western European colleagues who occasionally quote one or two Central European sites but did not seem to be aware of the full scope of the research happening in the region.

A second blog post on this project will follow soon featuring the first results of the citation network analysis, aimed at exploring this notion of unbalanced citation behaviour between the Eastern and Western researchers. Watch this space.

Iza Romanowska and Tom Brughmans