The Trees, the Forest, and the Passion for Prints

1. The Trees, the Forest, and the Passion for Prints Networks of Dutch Print Production, 1500-1750 Matthew Lincoln University of Maryland @matthewdlincoln July 22, 2015 Keystone Digital Humanities Conference
2. Art objects in the British Museum, by type @matthewdlincoln
3. “Sculptura in Æs”, from Johannes Stradanus’ Nova Reperta. Published by Philips Galle, c. 1588-1605. British Museum, London. What evidence can thousands of prints give us?
4. Designer “Sculptura in Æs”, from Johannes Stradanus’ Nova Reperta. Published by Philips Galle, c. 1588-1605. British Museum, London.
5. Engraver Designer “Sculptura in Æs”, from Johannes Stradanus’ Nova Reperta. Published by Philips Galle, c. 1588-1605. British Museum, London.
6. Publisher Designer Engraver “Sculptura in Æs”, from Johannes Stradanus’ Nova Reperta. Published by Philips Galle, c. 1588-1605. British Museum, London.
7. @matthewdlincoln
8. @matthewdlincoln
9. Peter Paul Rubens Schelte à Bolswert Gillis Hendricx designed @matthewdlincoln
10. British Museum collections LOD: collection.britishmuseum.org Between 1500-1750: •  49,306 prints •  3,592 nodes: distinct designers, printmakers, and publishers •  76,697 edges: connections inferred from co-participation in an object Mining the museum for data @matthewdlincoln
11. @matthewdlincoln Requisite Gephi mess:
12. @matthewdlincoln 1.  Create a set of subgraphs based on a rolling 10-year window: •  e.g. the 1640 subgraph contains only edges and nodes extant between 1635 and 1645 •  Edges (prints) exist between the start and end dates ascribed to an object o  Edges are unweighted to avoid biasing edge strength based on the number of surviving impressions (complicated!) •  Nodes (artists) exist during their life dates (also complicated!) 2.  For each subgraph, calculate network metrics at the global, regional/national, and individual scale Dynamic network analysis
13. @matthewdlincoln 1. Did the northern Netherlands adopt and continue the highly-centralized Antwerp print production model through the seventeenth century? OR… 2. Did rising Dutch prosperity instead support a more distributed network of local print markets? My question: centralized production
14. é More centralized ê More distributed @matthewdlincoln Graph centrality score
15. @matthewdlincoln é More centralized ê More distributed 0.00 0.05 0.10 0.15 0.20 1500 1550 1600 1650 1700 1750 year graph centrality score é More centralized ê More distributed @matthewdlincoln
16. 0.00 0.05 0.10 0.15 0.20 1500 1550 1600 1650 1700 1750 year graph centrality score @matthewdlincoln •  Rapid centralization around 1580-1600 •  Swift re-distribution within a generation, reverting to a low level by 1640s •  Economic contraction in 1670s did not lead to an immediate return of centralization
17. 0.00 0.05 0.10 0.15 0.20 1500 1550 1600 1650 1700 1750 year graph centrality score @matthewdlincoln Lucas van Leyden
18. 0.00 0.05 0.10 0.15 0.20 1500 1550 1600 1650 1700 1750 year graph centrality score @matthewdlincoln Hendrick Goltzius
19. 0.00 0.05 0.10 0.15 0.20 1500 1550 1600 1650 1700 1750 year graph centrality score @matthewdlincoln Claes Jansz. Visscher
20. 0.00 0.05 0.10 0.15 0.20 1500 1550 1600 1650 1700 1750 year graph centrality score @matthewdlincoln Nicolaes de Bruyn
21. 0.00 0.05 0.10 0.15 0.20 1500 1550 1600 1650 1700 1750 year graph centrality score @matthewdlincoln Abraham Blooteling
22. 0.00 0.05 0.10 0.15 0.20 1500 1550 1600 1650 1700 1750 year graph centrality score @matthewdlincoln Bernard Picart
23. centrality nodes edges 0.00 0.05 0.10 0.15 0.20 0 100 200 300 400 0 200 400 600 1500 1550 1600 1650 1700 1750 year value @matthewdlincoln
24. centrality nodes edges 0.00 0.05 0.10 0.15 0.20 0 100 200 300 400 0 200 400 600 1500 1550 1600 1650 1700 1750 year value @matthewdlincoln It’s not just the number of artists. It’s how they connect.
25. @matthewdlincoln BUT WAIT We want to avoid “just-so” stories! How do we know if this theoretical explanation makes sense?
26. Simulation time @matthewdlincoln IF our simulated network metrics appear similar to the observed network metrics from our dataset, THEN we can feel more confident about our proposed explanation. Let’s create a simulation w/ actor behavior we are proposing.
27. Erdos-Renyi: edges added totally at random Random graph generation Scale-Free: edges follow a power-law distribution @matthewdlincoln
28. 0.00 0.05 0.10 0.15 0.20 0.25 1500 1550 1600 1650 1700 1750 year graph centralization score model erdos−renyi scale−free3 @matthewdlincoln
29. 0.00 0.05 0.10 0.15 0.20 0.25 1500 1550 1600 1650 1700 1750 year graph centralization score model erdos−renyi scale−free3 @matthewdlincoln Printmakers needed expert collaborators
30. 0.00 0.05 0.10 0.15 0.20 0.25 1500 1550 1600 1650 1700 1750 year graph centralization score model erdos−renyi scale−free3 @matthewdlincoln Networks deserve metrics, not just viz. They also need simulation, not just speculation
31. Matthew Lincoln matthewlincoln.net @matthewdlincoln