Nodes and Edges!! Who knew!? Everything about the digital aspect of analyzing social (and other) relationships was new to me. As a novice, I carefully read through Scott Weingart’s “Demystifying Networks” article where he introduces the basic technical aspects of network analysis in an approachable manner. From explaining modalities to cautioning about methodology appropriation and algorithm limits, Weingart covers a lot of ground while encouraging beginners in the digital humanities to take risks and explore the possibilities within the constraints of current software. As he says, “Nothing worth discovering has ever been found in safe waters.”
The Weingart article lays a foundation for understanding how network analysis functions, but the case studies reveal the exciting potential of what is possible. Ruth and Sebastian Ahnert’s article, “Protestant Letter Networks in the Reign of Mary I: A Quantitative Approach,” examines the social relationships of the underground Protestant movement during the reign of Mary I using 289 letters, some of which were partially printed in Foxe’s Book of Martyrs. By looking at the original letters which included commendations, and information regarding financial sustainers (often female), their study reveals a robust social network along with the nuanced roles of those involved in the religious resistance. The authors use the concepts of “betweenness” (the shortest path to go through a node) and “eigenvector centrality” (which measures how well-connected a node is to hubs) to reveal the importance of some participants in this social network who were often written out of the historical narrative. Women as financial sustainers play a much more important role than previously recognized by historians.
Matthew Lincoln’s “Social Network Centralization Dynamics in Print Production in the Low Countries, 1550-1750,” takes a different approach using “time slices” to examine the printmaking industry in Northern Europe. Focusing on the robust Flemish and Danish print and engraving collections from the British Museum and the Rijksmuseum, Lincoln uses quantitative methods to analyze this representative sample of who was producing prints, along with where they were being made and disseminated for public consumption. Lincoln’s focus on network centralization is key to understanding the rise and fall of prolific and competitive printmaking in various cities, often associated with apprentices going into business and other social factors. Often art historians have only looked at printmakers who created original works, but Lincoln’s study reveals the importance of popular printmakers, like Jonas Suyderhoef, who copied original works for public consumption.
When read side by side, the Ahnert and Lincoln articles show how different methodologies and diverse approaches can be used to examine aspects of history that were previously inaccessible. The Ahnert article was particularly compelling for examining nuanced social interactions. I was struck by the concept of “node failures” and how the loss of one person/connection splinters communities. Ahnert’s findings assert,
“node failures can easily break a network up into isolated, non-communicating fragments….in a social or ecological network it could easily be caused by illness or death. What studies have shown is that one of the most effective ways to fragment a network into separate communities is to remove nodes or edges with the highest betweenness, a key measure of interconnectivity in the network.”
This concept should resonate with anyone who has lost someone in a similar role of holding a community/group together. But also, it shows how Mary I’s strategy of removing prominent members of the Protestant community could fragment the movement.
This module also included three online network analysis projects to examine. Because I dabble at being a Tudor history and visual culture nerd, Six Degrees of Francis Bacon was probably my favorite project to explore. I was really surprised that Sir Francis Drake only had two connections! It was really interesting to visualize the hubs of activity and connections around Francis Bacon, King James I & IV, Queen Elizabeth I, Robert Cecil, and others. It made me further appreciate the Gephy activity.
Which brings me to the technical activity. Gephy was both fascinating and frustrating at times. I had the initial issue with the known Java error. However, after that was fixed, I encountered an error with loading/opening the program. Every time it hit the “loading modules” portion of startup, the program closed. I did some troubleshooting on the forums and found where a few others encountered this. I checked my version of Windows and found that I’m running 64-bit and so I downloaded the 64-bit version of Java and repointed the software at it. I got another Java error, but this time when I clicked through it, the program opened.
After I was finally in, I started playing around with the sample data. Some of the features are not intuitive and if I wasn’t following the class tutorial, I would have been completely lost. As it was after a lot of trial and error, I finally started to get the hang of it. The first semblance of a network analysis data visualization for this dataset comes in the form of a circle of non-connected nodes (0 degree connections) with a center portion of connections, which looks like the “Great Eye of Sauron” from Lord of the Rings (above left). The center portion became readable when zoomed in, but was hard to decipher. Many of the brightly colored labeled nodes see here have degrees less than 10(above right and lower left). I started to make this easier to decipher by trying to use the “prevent overlap” feature in Gephy, but that only expanded the very dense data a little (see middle photo below). I then attempted to remove some of the lesser degrees to look at the stronger connections. This starts to be an approachable visualization, where hubs can potentially be found.
A close-up of the nodes from the dataset. The center portion after expanding by “preventing overlap.” The dataset after removing some of the lesser degree nodes.
I went a little further and removed the nodes with degrees less than one hundred to get to an easier to read (but very incomplete) visualization. Because there are often similar names, I chose to include occupations in the labels. This proved helpful in the below visualization that includes 3 Francis Bacons which are unconnected by any edges in this dataset/algorithm. The image below to the right has the nodes with the highest levels of degrees for this dataset.
This one has 3 Francis Bacons without any connections. These are the nodes with degrees between 198-565
After working with the dataset and names, it reminded me of Module 2 and learning OpenRefine. I couldn’t find where I saved the extensive edits on their spreadsheet regarding the names of royal women, but I edited this dataset to be similar and ended up with a few good visualizations. To start, I reset the filters and ended up with another “Great Eye of Sauron” for just the nodes marked as “female.”  Next, I removed the 0 degree nodes and looked at the center connections(lower middle image). I further reduced the degrees (eliminating 1 & 2) and the result is the image to the lower right. A striking difference with the male heavy data above is that none of the women in this dataset have a degree higher than 15. So, by eliminating the nodes with degrees below 100 in the earlier images, all “women” in the dataset were removed. Being able to look closer at these “female” connections make me wonder where the algorithm is pulling its edges from. For instance, why aren’t Katherine Parr and Queen Elizabeth I connected, since Elizabeth lived with Katherine for a time during her youth? And why are Elizabeth I and Anne Boleyn unconnected? Along with Mary I and Katherine of Aragon? I would have assumed that parent & child edges would be included, but that’s also an assumption on how the data is set up. I’d love to know more about what is driving the edges/algorithm in this instance, but I think it’s also a good reminder to be mindful and critical of how data is interpreted and shown through network analysis.
An image of just the nodes marked as “female” in this dataset. An initial look at the center connections. A more readable version of the center connected nodes, but also showing how some connections are missing.
There are so many exciting applications of this methodology. I’m really glad that I decided to push myself into doing another optional module and had the opportunity to explore this new realm of knowledge.
 Ruth Ahnert and Sebastian E. Ahnert, “Protestant Letter Networks in the Reign of Mary I: A Quantitative Approach,” ELH 82.1 (2015).
 Matthew Lincoln, “Social Network Centralization Dynamics in Print Production in the Low Countries, 1550-1750,” International Journal for Digital Art History 2 (2016).
 P. 27
 I recognize that gender is a spectrum and includes more than the binary terms “male” and “female.” Unfortunately, historic records are often limited, as is the case with this dataset. Therefore, I’m using the binary terms prevalent in this time.