Peripatetic Networks

19/05/2021

The ‘Networking Archives’ Team

Worked is based on a larger project, thanks to (clockwise from left): Howard Hotson (Principal Investigator), Miranda Lewis, Matthew Wilcoxson, Arno Bosse, Philip Beeley, Ruth Ahnert (Co-Investigator), Sebastian Ahnert (Co-Investigator), Esther van Raamsdonk (http://networkingarchives.org/project-team/)

Historical and spatial networks

Interest in Historical Network Analysis (HNR), also seen interest in understanding spatial and temporal dimensions of networks
Spatial information in networks used to map letter origins or paths
Or used as a weight function in a network
Richard White, Spatial History Lab:

I don’t want to be so simplistic as to say that if space is the question then movement is the answer, but I fear that I am nearly that simple. (Richard White,‘What is Spatial History?’, 2010)

I propose a number of ways in which we can incorporate movement into networks

In recent years the field of historical network analysis has grown considerably and atthe samee time there’s also been an interest in understanding spatial and temporal dimensions of historica, networks. Spatial information in networks is often used to map letter origins, or drawn paths between origins and destinations, or used as a weight function in a network.

But space is also about movement. To quote Richard White of the Spatial history lab in Stanford, who says:

But this definition of space as mostly been about movement has not so far made its way to historical network analysis. So I would like to propose a number of ways in which we can incorporate movement into networks and present a number of experiments and case studies on how this movement can tell us something about centrality and reasons for it.

The Data Used

Early Modern Letters Online

A union of c.100 catalogues brought together to study the Republic of Letters (1500-1800 but focus in 17th century)
Letters from across Europe, focus in particular on Dutch correspondence

State Papers Online

Digitised Calendars of scans of the State Papers from Britain and Ireland, Tudor and Stuart Periods (1509 - 1714)
SP might be thought of as the principal and working library for the Parliamentary executive
Both domestic and foreign papers

First a couple of words on the data used. The project that I’m working on is conducting network analysis and other data analysis techniques on a large dataset of correspondence. The first is a dataset called ‘Early Modern Letters Online’, which is an online database consisting of a hundred or so separate letter archives, brough together to study the republic of letters. It spans 1500-1800 but with a focus in the 17th century. There are letters from right across Europe, but with a focus on the Dutch republic.

The seocnd dataset is the state papers online. These are digitsed scans of the calendars of state papers for britain and ireland, for the tudor and stuart periods, which runs from 1509 to 1714. The SP might be thought of as the principal and working library for the Parliamentary executive, and contains both domestic and foreign letters.

Geographic data

Out of 470,000 records, 295,000 have sets of coordinates for letter origin; about 57,000 also with destination information
Letters mostly in Britain, Ireland, Continental Europe but also America, India etc.
Information used to map the geography of the letter archives:
- EMLO is particularly European, with a concentration in the Low Countries
- State Papers Stuart more English: mostly domestic state papers
- Tudor slightly more European-focus, also Dublin and Edinburgh

Movement in early modern Europe

17th century was a time of movement for many
- Growth of the postal system meant letters could be carried across increasingly large distances at less cost
- Elites in Britain conducted ‘Grand Tours’ of the continent
- Diplomatic missions could mean travelling across Europe for months at a time
- Also those displaced by the 30 years war, which involved much of continental Europe and resulted in armies, refugees constantly moving.

These maps only show the movement of letters, but people themselves also moved around at the same time. The seventeenth century was a time for increased movement for many in society. There were some technological advances: letters could be sent across long distances at lower costs. Elites in Britain conducted grand tours of the continent, travelling across Italy and other places. Diplomatic missions could mean travelling across Europe for months at a time. And there were those displaced by war: the Thirty Years’ War from 1618 to 1648 invovled much of continental Europe and involved armies moved around.

Travel didn’t stop these people from writing: when abroad it was very normal to continue a correspondence as when at home or even start new ones, often leaving instructions in a letter letting the recipient know where they should send the next one so that it would be waiting for them when they arrived. This means that much of the record of this travel can be found in the letter records.

Tracing Movement

Some of this movement can be traced through correspondence data.
We calculate an ‘itinerary’ for each person in the dataset
Then measured the distance of each route, as the crow flies

Tracing Movement

18 Oct 1635: Dury, John, 1596-1680 (The Hague, South Holland, Netherlands) to Roe, Thomas (Sir), 1581-1644

Tracing Movement

7 Apr 1636: Dury, John, 1596-1680 (Amsterdam, North Holland, (United Provinces) Netherlands) to Godemann, Caspar, fl. 1631-1643 (London, England)

Tracing Movement

29 May 1636: Dury, John, 1596-1680 (Norrköping, Östergötland, Sweden) to Roe, Thomas (Sir), 1581-1644 (London, England)

Tracing Movement

3 Jun 1636: Dury, John, 1596-1680 (Stockholm, Stockholm Municipality, Sweden) to Hartlib, Samuel, 1600-1662 (London, England)

Results

Transform this into a ‘simple features’ spatial object of linestrings
Allows all ‘itineraries’ to be mapped and their distances measured.
Captures a variety of movement such as Thomas Windebank and Thomas Roe

Tracing Movement

Mobility and Networks

Next I was interesting in investigating the relationship between this mobility and network centrality measures.
Calculated Spearman correlations between distance rank and centrality ranks
Moderate correlation can be found (I consider this a non-result)

Mobility over time

Created linestrings for each year for each person, and measured these to produce ‘mobility charts’
Used a rolling sum of the previous three years to smooth out year-by-year changes
Resulting data can be used to find interesting peaks - periods in individuals’ lives where they were particularly mobile

To further investgate the relationship, a longitudinal study was carried out. This was done because the data is longitudinal: we have sparse records over a long period of time (I focused on 100 years for this study). This means that an individual might travel extensively during one period and have many connections in another one entirely, resulting in a false correlation between the two. First I calculated a rolling sum of travel distance for the previous 3 years. The resulting data can be used to find interesting peaks, for example by looking for observations which rise a certain number of standard deviations above a local mean: this can find periods in individuals’ lives where they were particularly mobile. Using the example here again of John Dury, whose mobility comes in two clear peaks, between 1636 and 1639, and again between 1650-1655. These do correspond with known points where Dury was abroad, on diplomatic or ecclesiastical missions.

Mobility and network metrics

Allows us to assess effects of mobility and travel on networks in 17th century
Compared network and mobility scores
In Dury for example, increased degree centrality follows closely behind increased mobility: travel was a way of establishing new contacts.

Looking systematically for movement influencing centrality scores

Looked for short periods with high correlation between mobility and degree
Calculated Pearson’s values for 4-year rolling window
Also calculate with 2-year lag and lead (looking for cases where a burst of travel precipitated an expanding network)

Looking closer at the relationship between degree and mobility

To estimate the effect of travel on networks, wrote a function which removes edges from a selected period, recalculates centrality scores for a given node, and lists individuals who they had contact with again after the travel period ended.
Function compares the change in ranks if a period of travel had not happened
For Dury between 1636-1639, many of his important connections were during this tour of the Netherlands and Sweden.

Contacts up to that point: 82 Contacts during voyage: 23 (14 for the first time) 6 were a contact again after the voyage: St Amand, Joseph (300756) Groot, Hugo de (900019) Dingley, John (300342) Rusdorf, Johann Joachim von (300175) Laud, William (300290) Oxenstierna, Axel (Count) (300658)

CHANGES degree rank before: 26 degree_rank_after: 28 Change: -2 Betweenness rank before: 21 Betweenness rank after: 24 Change: -3 Eigen centrality rank before: 18 Eigen centrality rank after: 24 Change: -6

As already hinted at, the hypothesis is that travel facilitated Dury’s network in some way: that moving around actually increased his contacts. To look more closely at this I created a function that takes a set of dates and produces a report of the changes before and after that period. With a single command, it lists all places visited, people met, people who that person stayed in contact with after the trip, and changes in overall network centrality scores if this period of correspondence was removed from the network. As you can see with Dury, between that period Dury was in contact with 23 people, 14 for the first time, and 6 who he stayed in contact with afterwards. This suggests that Dury’s voyage increased his centrality to the network in which he was invovled. Intriuguingly, it’s actually Dury’s eigenvector centrality score that is most affected by his travel: had he not made the journey, his rank falls by 6 places. This is perhaps because it’s only when he is away from home that he has the need to contact central, important nodes.

Conclusions

In historical networks understanding networks with respect to movement is as important as a typical spatial perspective
Comparing the two has shown:
- Movement does not necessarily correlate with network centralities but it can do
- Rolling correlations help to understand periods where the two match up