How We Use Twitter Part 1: #MedievalTwitter Roundup from December 23, 2016

On December 23, at a time when the President Elect can turn the world on its head with a tweet, or a simple sentence about arms races, I thought I’d see what we medieval folk were talking about on Twitter…

I’ve been playing with the NodeXL network analysis tool for Excel, both to get back into social network analysis and to explore some of the capabilities of the platform. I’d only begun to test it in the aftermath of the Great Crusades Debate of February 2015 (for which I still have all the data, and will be blogging that soon-ish, or finding a place to publish it). Anyway, I wanted to see what kind of discussions we had been having before Christmas, and to discover which conversation topics generated the most interactions. In a sense, then, this can be seen as followup to my earlier post, Don’t Oversell Twitter, this time with some hard data and perhaps a more nuanced interpretation of such data. [Also, please note, before you get all fired up, that all the information captured here is public data.]

Parameters and Terms

The program I like best  for this kind of analysis is NodeXL. NodeXL has several specialties, but among its best features is its ability to analyze Twitter networks.  I decided to take a look at two hashtags, at 1200hrs and 1500hrs on December 23 (I didn’t finish writing this post till today), and see how they traveled about the platform, in terms of interactions.

There are a number of things to remember before diving in. NodeXL doesn’t measure impressions, that is the number of times someone saw the tweet, or engagements, that is the number of times someone interacted with the tweet (these are Twitter’s own definitions). This needs to be remembered when looking at the data captured here on interactions. Relevant terminology, for those not familiar with network analysis lingo:

Vertex: Also known as “nodes,” these are the individual items that make up the network. In this case, the nodes are individuals, or individual Twitter accounts.

Edge: This is the connection between two nodes, or vertices. Most network analysis programs can analyze edges in several ways. They can display edges simply to show that there is a connection, without specifying what kind it is: this is called an “undirected network.” If you change the settings to display who is doing the action to or toward whom, that is a “directed network.”

Directed/Undirected: As above. The graphs here are directed, because I want to see who is performing the action to whom. An arrow to a node indicates that the node has been mentioned, quoted, replied to, or retweeted.

Centrality: For a good overview of “centrality,” see UC Riverside’s page on it, and Peter Hoff’s easy-to-grasp PowerPoint from the University of Washington.

Betweenness Centrality:  This is the first of several common ways of testing how important various nodes are in the network. “Betweenness” indicates how often a node lies on the path to other nodes–in other words, it’s a way of figuring out who controls access, or serves as an access point, for information, and hence who has the most influence in a network.

Closeness Centrality: Closeness indicates how far a vertex (node) is from other nodes. In other words, it measures who knows the most people, or who connects with the most people. It doesn’t tell us how influential that person is–they may just be sociable.

Eigenvector Centrality: Eigenvector is a more discerning way of calculating In/Out Degree (how many links does a node receive or give). Not all nodes are equal–so, if you’re a node, the Eigenvector measurement calculates the relative importance of the nodes that you connect to.

With these terms in mind, let’s look at the data.


The first one is #medievaltwitter itself, as of about 1500 hours, December 23, 2016. Full data, including the worksheets, can be found at the NodeXL Gallery. I used the Harel-Koren Fast Multiscale layout because it shows more clearly the relationship between nodes (vertices) within groups. Note that this search was only for the hashtag #medievaltwitter, and not #medieval (I’ll run a search on both terms in my next post or two, as that will throw up a much richer network). For December 17 to December 23, the coverage allowed by Twitter, 454 users participated in the hashtag. Let’s look at the top five groups:

A couple things stand out from this, at least to me. First, the nodes with the highest “betweenness” centrality–those whose posts got the most interaction on Twitter–were nodes that shared pictures.

Group 1 (dark blue, 99 nodes) and Group 2 (light blue, 53 nodes) are centered on two individuals: Francois Soyer, who tweeted 13 times in the period captured by NodeXL, and Emily Steiner, who tweeted 4 times. In both cases, the author tweets that received major interactions had to do with art (which is what Professor Soyer and Professor Steiner often tweet about, much to our collective edification). Professor Steiner had one tweet that went very active, on a 15th-century Brabantese ‘Crib of the infant Jesus,’ while Professor Soyer had several (not including his own retweets).

Art was again the main topic of Group 4 (light green,35 nodes, cluster highlighted in red, right), centered on Rory Naismith’s tweet of a 12th-century Scandinavian chess piece.

So a basic read of #medievaltwitter is that we’re most enthusiastic about art and manuscript images–in other words, we are interested in the warp and weft of our trade.

This is where Group 3, dark green, 51 nodes, is very interesting. First, it has no fixed center (an advantage of the Harel-Koren layout, you can see this), and it centers entirely on issues of the profession and activism. In other words, this is where you will find people who tend to tweet about political, professional, and activist issues. There are two major “poles”: the Society for Medieval Feminist Scholarship’s tweet about the Trans Travel Fund for conference travel and David Perry’s tweet about the Scottish mosque vandalized with “Deus vult” and “Saracen Go Home” (weird and disgusting, but I haven’t seen any further update on this).

These received 19 and 14 retweets, respectively–not nearly as much interaction as art and manuscript images.

Between these two clusters are four nodes that connect the two: Rick Godden and Kathleen Kennedy each possess high Betweenness Centrality ratings, followed in much lesser degree by Matt Gabriele and T. S. Wingard.

Group 5, the last group I’m going to interrogate here, was centered around tweets from the University of Lancaster about their new lectureship in medieval history and Levi Roach, at Exeter (I highly recommend you follow him) retweeting that and tweeting other things. Group 5 also includes tweets from a number of people in my research areas, including De Re Militari, Andrew Ayton, and Iain MacInnes.

Just to be clear, I’m that solitary box down in the corner, the lonely nodes who had no interaction with folks. Or barely any. I was inquiring as to medieval commentaries and interpretations on the Book of Joshua, but got only one response. Oh well…

It should be noted that a heck of a lot of “medieval twitter” isn’t actually included in the hashtag. Many institutions that tweet medieval things don’t include any “medieval” or “medievaltwitter” hashtag in them. Further, who is the most influential tweeter depends entirely on what metrics you use to calculate it (refer to the “Top Items” column in the spreadsheet). If you look at “betweenness,” fjsoyer and piersatpenn are the top two. Yet by “closeness” calculations history_geek has the most immediate reach, because she has the most followers-but not necessarily the most influence. Which is why, when you calculate influence according to “Eigenvector,” fjsoyer appears the most influential #medievaltwitter user in this week.

So, what can we learn from this incomplete picture? Generally what I said above–most of us like tweeting and retweeting pictures and sources that help us teach and research more effectively. Job announcements, politics, activism, and such are far less prominent topics of conversation, at least under a #medievaltwitter hashtag. Personally, I don’t think that’s a problem, but others might.


The second network is a hashtag that my colleague at Dominican University, David Perry, started using: #NewBloodLibel (full data at NodeXL Gallery). It refers to Fox News and Breitbart claiming that a Lancaster, PA school canceled a production of “A Christmas Carol” because of the moral objections of a Jewish family. Complete fabrication, but the family has since fled Lancaster for the time being. That this could happen in America in the twenty-first century, right as the Holocaust generation is dying out, is especially disturbing. (Of course, the vectors of this are extremely complicated, both inside and outside of Israel itself–among other factors, it depends on your equating of opposition to Israeli settlements with Antisemitism. Not touching that with a 39.5-foot pole). In any case, it doesn’t seem to be catching on, if the metrics and visualization are anything to go by. But with it seeming more likely  that the Nazis will be marching in Montana after all, I have a feeling we’ll be needing this hashtag more and more in the new year. Feels like the 1930s, all of a sudden…

To be continued…