Visualizing Networks of Artistic Ideas in History Paintings in the Seventeenth-Century Netherlands

Existing studies have demonstrated the potential of network analysis in extracting and visualizing complex information from large historical datasets (Larson 2013; Sun 2016). Art history, however, still dwells in the early stages of exploring how network methodologies can benefit its scholarship. The only applications of network analysis in art historical research often equate network to social network and borrows concepts and theories from the social sciences (Kok 2013; Brosen et al. 2016; Lincoln 2016, 2017). Digital art history calls for new concepts tailored to artistic networks in order to account for the developments of art in all its complexity.
This research conceptualizes a novel art historical network of ideas inscribed in the iconography, connecting artists not through social ties but shared subject matters. Through visualizing this conceptual network, this study tries to answer the following questions: how are artists connected through their choice of subject matters? Have the patterns of such connections changed over time? And more importantly, is the choice of subject matter related to the choice of painting style, location of residence, and timing of entering the market? To answer these questions, this study draws on large digital collections of Netherlandish paintings to construct a dynamic network model of iconography. This research foregrounds history paintings produced in the Low Countries between 1575 and 1700, an established genre which witnessed dramatic iconographical changes during this period (Sluijter 1991). The results of this study demonstrate the spatial and temporal evolution of the artists’ network of ideas, revealing the changes in the structures of artistic interactions and the diffusion patterns of subject matters within the artist community in different cities.

Data
This research takes advantage of the large online collection of paintings of the Netherlands Institute for Art History (
RKDimages).

https://rkd.nl/en/explore/images

This database describes over 150,000 paintings within the scope of this study. In the absence of a mature image recognition technology that is able to discern a great variety of complicated subjects of the paintings, this research uses the iconographical notation system ICONCLASS, a hierarchically ordered system of unique alphanumeric classification codes with which most images in the RKD have been processed and tagged.

Iconclass is a classification system designed for art and iconography. It is the most widely accepted scientific tool for the description and retrieval of subjects represented in images (works of art, book illustrations, reproductions, photographs, etc.) and is used by museums and art institutions around the world, such as the Rijksmuseum. More details see:
http://www.iconclass.nl/

History paintings in the RKD database are distinctively labeled with ICONCLASS often referring to a specific scene or story, like
Adoration of shepherds (73B25) and
Diana discovers Callisto’s pregnancy (92C35211). I further link the
RKDImages to the comprehensive biographical database,
ECARTICO, to supplement the paintings with the biographical information of the artists.

ECARTICO is a comprehensive collection of structured biographical data concerning painters, engravers, printers, book sellers, gold- and silversmiths and others involved in the ‘cultural industries’ of the Low Countries in the sixteenth and seventeenth centuries:
http://www.vondel.humanities.uva.nl/ecartico/

The active period of the artists from
ECARTICO is used to introduce the dimension of time to examine the evolution of popular subjects and to evaluate the strength of connections among generations of artists.

Method
The network of ideas draws a connection between two artists through a mutual subject matter; the strength of the tie between them is measured by the total number of iconographical overlaps in their oeuvre. The record of a painting in the
RKDImages identifies a link between the painter (if attributed) and the subject (labeled with ICONCLASS). One painter can link to many subjects as he painted various scenes and one subject matter can tie to numerous painters who depicted it. Putting all the works of art and artists together, a two-mode network is constructed, which consists of two types of nodes: artists and subject matters.

The two-mode network of persons and subjects are then collapsed into two one-mode networks by converting one of the node types to an edge or link between nodes of the other type (Watts 2004; Hanneman and Riddle 2005; Knappett 2011). For artists who picked a particular subject matter, the subject becomes the conceptual edge connecting the two artist nodes, while a single artist who depicted two stories would be the edge between those two subject matters. In other words, the two-mode, artist-subject network is projected into 1) an artist-artist network, in which the link is constructed by the subject matters artists mutually depicted; and 2) a subject-subject network in which the link is drawn by the artists who painted both themes. I resort to graph theory to examine the nature and the properties of the network of ideas. Among the measurements, modularity is particularly relevant to art history as it helps us to identify clusters of artists that are more likely to paint the similar subject matter or groups of subjects that are more often co-existent in a single painter’s oeuvre. Modularity can help understand whether or not the thematic choice is related to the choices of other aspects of their work and life.

Results and Discussion

Changes in the network of ideas over time
Artists are grouped by the year they entered the market following Rasterhoff (2017)’s periodization of the art market in the early modern Netherlands and the network of ideas is constructed among artists in each period (Table 1).

Rasterhoff (2017) introduces the three periods of the painting industry and art market in the early modern Dutch Republic: 1580-1610 as a period of transition; 1610-1650 as an expanding market; and finally, 1650-1800 as mature market for paintings and other cultural goods.
A simplified version of the networks is visualized in Figure 1, showing the iconographies shared by at least three artists. Nodes in Figure 1 are sized by weighted degrees and are colored by the modularity classes.

Table 1. Metrics of the network of subject matter by time period

# of edges

#of Iconclasses

Avg. Degree

Degree centrality

Avg. Clustering coeff.

Avg. path length

Modularity

1575-1610

89

35

5.09

0.53

0.75

2.26

0.39

1610-1650

237

75

6.32

0.62

0.64

1.90

0.38

1650-1700

380

57

14.58

0.40

0.73

1.97

0.33

The structure of networks shown in Table 1 indicates an expansion in popular subject matters between 1610-1650, coinciding with higher interconnectivity seen in the network of artists. Astonishingly, although a greater variety of subject matters have been explored and popularized by the generation of masters who entered the market between 1610 and 1650, their successors fell back to a shrunken pool of subject matters and restrained their repertoire, which is reflected in the reduction of the number of nodes. This observation of the popularity of a limited number of subjects and the virtual absence of others in the evolution of the network of subject matters verifies the limited repertoire art historians suggested (Sluijter 2000; Bok 2001). Furthermore, a structural change is observed in the evolution of the network across time periods: the connections among the subjects depicted by painters entering the market after 1650 are more evenly distributed (Fig. 1c) and the network is less centralized, meaning the painter was likely to pick among the same pool of subjects instead of venturing into new iconographies.
(a) 1575-1610

(b) 1610-1650

(c) 1650-1700

Figure 1: Network of subject matters by time periods
Remarkably, the clusters of subjects (Fig.1, marked by colors) do not speak to the categories modern art historians use such as mythology, Old and New Testament. Rather, one can find a mix of all three in one cluster. The clusters in Figure 1 relate more closely to the visual resemblance of the scenes and the skills required to paint them. For instance, both the
Destruction of mankind (Old Testament) and
Marriage of Peleus and Thetis (Mythology) from the same cluster, depict scenes that are usually composed of a large group of nude figures in an outdoor setting.

Geographical variation of taste and connection
The network of ideas infers a certain degree of locality, as observed in the size and structure of the network constructed by artists who mainly worked in the following artistic centers: Amsterdam, Haarlem, and Utrecht (Table 2). It means the concept of ‘local schools’ did exist before 1610 but gradually dissolved in the following decades when the market was more integrated and painters living in different cities could keep abreast of each other reflected in their choices of subject matters.
Table 2. Metrics of the network of artists by location

# of edges

#of painters

Degree centrality

Avg. Clustering coeff.

Avg. path length

Modularity

Amsterdam
140
43
0.48
0.59
2.24
0.17

Haarlem
29
16
0.87
0.68
1.93
0.08

Utrecht
43
18
0.68
0.64
1.87
0.10

(a) Amsterdam

(b) Haarlem

(c) Utrecht

Figure 2: Network of subject matters by city
The networks of ideas in Amsterdam, Haarlem, and Utrecht demonstrate unique characteristics: Amsterdam had higher modularity which loosely corresponds to different styles co-existing at the same time; Haarlem had a centralized network and a schism between generations, and Utrecht enjoyed the most densely woven networks of both artist and subject matters. The choice of subject matters shows certain preferences in each city, with Amsterdam venturing more into the new subjects popularized by innovators from Rembrandt’s teacher, Pieter Lastman, to Rembrandt, Haarlem demonstrating the legacy of Mannerism, and Utrecht embracing both old and new fashion in its oeuvre.
Amsterdam has the largest network among the three artistic centers. The high average path length and the relatively low clustering coefficient indicate a more hierarchical structure around key painters and the flow of artistic knowledge is likely to have trickled down from the great masters to their circle, then to the lower segments, instead of diffusing among the minor masters (Fig.2a). Amsterdam’s network is also the least centralized among the three (Table 2), suggesting that several masters formed their relatively independent circle of interaction.

Bibliography

Bok, M.J. (2001). "The Rise of Amsterdam as a cultural centre: the market for paintings, 1580-1680", in: P. O'Brien, et al. (red.),

Urban Achievement in Early Modern Europe: Golden Ages in Antwerp, Amsterdam and London

.
Cambridge: Cambridge University Press,186-209.

Brosens, K, et al. (2016). "MapTap and Cornelia. Slow digital art history and formal art historical social network research." 
Zeitschrift für Kunstgeschichte 79-3: 315-330.

Hanneman, R. and Riddle, M. (2005).
Introduction to social network methods. Riverside, CA: University of California, Riverside.

Kok, E.
(2013).

Culturele ondernemers in de Gouden Eeuw: de artistieke en sociaal-economische strategieën van Jacob Backer, Govert Flinck, Ferdinand Bol en Joachim van Sandrart
. PhD diss., Universiteit van Amsterdam.

Knappett, C. (2011)
An archaeology of interaction: network perspectives on material culture and society. Oxford: Oxford University Press.

Larson, K.
(2013)."A Network Approach to Hellenistic Sculptural Production." 
Journal of Mediterranean Archaeology
26, no. 2: 235-259.

Lincoln, M. (2017). "Continuity and Disruption in European Networks of Print Production, 1550-1750." 
Artl@s Bulletin 6, no. 3: Article 2.

Lincoln, M. (2016). "Social Network Centralization Dynamics in Print Production in the Low Countries, 1550-1750."
International Journal for Digital Art History, no. 2.

Sluijter, E.J.
(2000)
Seductress of sight: studies in Dutch art of the Golden Age
. Zwolle: Waanders.

Sluijter, E.J. (1991).
"Didactic and Disguised Meanings? Several Seventeenth-Century Texts on Painting and the iconological Approach to Dutch Paintings of this Period
". In: D. Freedberg en J. de Vries (red.),
Art in history. History in art. Studies in seventeenth-century Dutch culture. Santa Monica, CA: Getty Center for Art History and the Humanities,175-207.

Sun, Y.
(2016). "The Social Network behind an Architectural Style over Space and Time." In 
Space and Time Visualisation
. New York: Springer International Publishing, 79-96.

Rasterhoff, C. (2017). 
Painting and Publishing as Cultural Industries: The Fabric of Creativity in the Dutch Republic, 1580-1800. Amsterdam: Amsterdam University Press.

Watts, D. (2004).
Six degrees: The science of a connected age. New York: WW Norton & Company.