The network of iconography draws a connection between two artists through mutual choices of subject matter (the term ‘iconography’ in this essay is used interchangeably with ‘subject matter’ to indicate the content of images as a whole, without interpreting possible meanings).12 A link in this network between two painters is measured by the total number of overlaps in the subject matters found in their oeuvres. Constructing such a network generally requires a source that meets two criteria: first, it should offer a sizable pool of artworks attributed to a significant number of artists; and second, it should identify subject matter in artworks with standardised references, free from consistency issues and amenable to categorisation and comparison. These rather daunting prerequisites may have thwarted previous attempts to build a network of iconography, given that the image recognition technology is not yet mature enough to detect subjects of visual arts, were it not for the digitalised archives in the Netherlands Institute of Art History (rkd), which offers an opportunity to realise such a network, albeit within a limited scope.

The rkd Images database, the online version of the rkd’s visual collections, presents a comprehensive digitised collection of Dutch art. It features broad and deep collections of photographs and reproductions of paintings by Dutch and Flemish artists, containing more than 100,000 unique paintings (including copies of known, attributed paintings), drawings, and sketches produced between 1550 and 1750.13 While the rkd collection has remarkably rich holdings, it cannot be regarded as perfectly representative of the full range of paintings produced in the seventeenth century. For example, it under-represents low-quality paintings as most did not survive.14 The rkd’s own collection history may exacerbate this survival bias, as the collection originated in the personal photo archives of early twentieth-century art experts and collectors, which introduced a selection bias that favours high-quality works of famous seventeenth-century Dutch artists. That said, studying networks of artists and iconographies requires artistic originality and attribution. These two prerequisites rule out low-quality copies or cheap paintings without identified painters missing from rkd Images, meaning that a carefully crafted research focus can alleviate – though not entirely eliminate – biases from the database. Despite some limitations, then, rkd Images, with its sheer volume, wide coverage, and diverse sources, is viewed as one of the most comprehensive repositories for surviving and/or identified Netherlandish paintings.

Besides images, the rkd releases curated data on its collection.15 It employs a hierarchical notation system known as Iconclass to describe subject matters and label works of art.16 A shared Iconclass tag denotes that paintings depict the same subject matter, thus drawing a thematic connection between these works of art. As such, rkd Images offers a unique opportunity to create a network of iconography at scale. Admittedly, this approach favours artworks falling into the broad category of history painting that depict moments in narrative stories in the Bible, mythology, history, and literature. These paintings are arguably the most researched and best documented in this database: they are the most thoroughly labelled with Iconclass tags and often have explicit references to scenes or stories, such as Diana discovers Callisto’s pregnancy (92C35211). Other genres like landscape and still life are often bereft of Iconclass labels, as they do not have an apparent iconographic reference to scenes or stories. It is worth noting that the Iconclass labelling should not be confused with the keywords encoded in the same dataset. Iconclass often compresses complex scenes into a succinct code, like the Adoration of the Shepherds (73B25), whereas keywords are more descriptive, listing for example genres, scenes, and motifs.17 Furthermore, the Iconclass classification system with its standard vocabulary is used by museums and art institutions worldwide, thereby suffering less from uniformity and consistency issues than the keywords.18 It must also be noted that the genre of history painting underwent dramatic iconographical changes in the seventeenth century. Fig. 1 shows the number of new Iconclass codes introduced each year, suggesting that novel subject matters were concentrated in history paintings depicting scenes from the Old and New Testament (yellow and bright green) as well as from classical mythology (light blue).

To take full advantage of the richness of the dataset and bypass its inherent limitations, this essay will restrict itself to history paintings (including landscapes with small-figured histories), prints, drawings, and oil sketches concerning stories from the Bible, mythology, classical history, and literature by artists active in the Dutch Republic.19 This limited scope guarantees the best coverage and high-quality Iconclass labelling in rkd Images. This research takes into account not only paintings but also drawings and oil sketches, because such preparatory sketches and studies also point to the awareness of an idea or a subject among artists, regardless of whether that idea was eventually executed in a painting or not. To guarantee the consistency of the Iconclass labelling, I manually processed the relevant labels to mitigate any possible flaws in the database.20 I then linked rkd Images to the ecartico database to add biographical information about the artists, in order to understand if artists could have seen each other’s work through their social connections or because they resided in the same city or neighbourhood.21 In particular, I used artists’ active period and location as listed in ecartico to introduce a temporal dimension, tracing the evolution of popular subjects among generations of artists in the Dutch Republic. Admittedly, the dataset inevitably imposes a Dutch-centric view of artistic connections among Dutch painters and artworks, and overlooks the fact that painters sometimes drew inspiration from visual sources from abroad or from non-visual ones such as literature.22 The development of Dutch art was in fact more intertwined with the wider European art world than this article may suggest. It is hoped, however, that the methodology introduced in this essay can be applied to more comprehensive sources and datasets when these become available in the future.

I took advantage of network theories and analytics programs such as rstudio and Gephi to translate the vast visual database into clouds of connections, visualising a network of artistic ideas and iconographies. This section will briefly explain how the iconography network is developed conceptually, leaving technical measures to Appendix i. The record of an attributed painting in rkd Images identifies a link between the painter and the subject matter (labelled with Iconclass). One painter can link to many subjects as he painted various scenes, and one subject matter can be tied to numerous painters who picked it up. Putting them together, I created a two-mode network with two types of nodes: artists and subject matters. Fig. 2a gives an example of such a two-mode network of artists (red) and subjects (blue), including around fifty scenes in mythology like Venus and Adonis and Diana and Callisto, both popular themes painted by many artists. Painters who created a variety of mythological scenes stand out with numerous connections (larger in size).

A closer look reveals how and why painters were connected artistically. Fig. 2b shows a surprisingly high degree of separation between the painters who painted Diana bathing with her nymphs and Diana hunting with her nymphs. Relatively few artists painted both subjects (highlighted in yellow), even though the two themes are closely related. The Dutch Republic seems to have had a long pictorial tradition of Diana’s bathing scene, as we find artists ranging from Cornelis van Haarlem in the early seventeenth century to painters of the next generations, such as Rembrandt, and later Arnold Houbraken and Gerard Hoet. This phenomenon might be explained by the iconographical differences and the distinct painting skills required: Diana bathing with her nymphs presents many female nudes, whereas figures in Diana hunting are clothed; and painting bare skin presumably requires different knowledge and a different cultural, religious, and moral environment.23

Although the two-mode network offers interesting insights, it does not demonstrate how artists were connected directly through shared subject matters. Nor can it be analysed by existing network theories, which are designed exclusively for one-mode networks that include either artists or subject matters, but not both.24 To create such a one-mode network of artists, we must convert (or ‘project’, in network parlance) the two-mode network shown in fig. 2 by turning the shared subject matters into a conceptual link between two artists (fig. 3; see for more details Appendix i). Likewise, a one-mode network of subject matters is generated by linking two subject matters by artists who depicted both. This conversion or ‘projection’ process eventually results in two one-mode networks: one of the artists connected by mutual subject matters; and one of the subject matters linked via the artists who painted them. These two networks allow us to examine the artistic connections among artists through subject matters, and to observe changes in thematic fashion over time.

This section traces the changes over time in the network of iconography, grouping artists by the time they entered the market. To do so, I have relied on the three phases in the development of the Dutch art market as identified by Claartje Rasterhoff: a transition period from a traditional society to a modern economy between 1580 and 1610, followed by a rapid expansion over the next forty years, until the market reached a state of maturity after 1650.25 The two networks of artists and subject matters link artists who entered the market during each of these three periods through at least two mutual subject matters in their oeuvre to avoid coincidences. Tab. 1 shows the structure of these two networks during different stages.

Tab. 1 shows that the network of artists tripled in size between 1610 and 1650, but shrunk significantly afterwards, which, given that it accords with existing scholarship, testifies to the validity and reliability of this method.26 The number of links (which represent the connections among artists) increased fivefold between 1610 and 1650, signifying a structural change. The network of subject matters, too, witnessed a rapid growth in the variety of Iconclass codes from 1610 to 1650. Yet despite the great variety of subject matters explored and popularised by the artists during the second period, their successors relied on a much smaller pool of subjects, which were surprisingly more densely connected than in the preceding period. The fact that the network of artists and that of subject matters diverge in their network density and structure suggests a shift in the thematic choices of artists (see Appendix i for more network measures). In what follows, I will discuss the characteristics of the network of iconography within each of the three periods.

Fig. 4 visualises the iconographic networks of artists entering the market between 1575 and 1610. Mannerist giants like Cornelis Cornelisz. in Haarlem and Abraham Bloemaert in Utrecht occupy key positions in this network. They not only enjoy a close iconographic bond but also reach painters in relatively peripheral positions. This network of painters in the first period is highly centralised, as it is dominated by a few painters working in a similar style and depicting similar subject matters while residing in different cities.

The colours in fig. 4 mark the clusters of artists who are more likely to paint similar subject matters within the group as opposed to those outside of it. These colour-coded clusters represent the so-called modularity classes or modules in network analysis that detect densely connected clusters within a network (see Appendix i). For instance, fig. 4 shows that the painters in blue shared more subject matters among themselves, and as such were separated from the red group, which was dominated by the aforementioned Mannerists. A closer look reveals that the blue cluster consists mainly of Amsterdam painters, led by Pieter Lastman and David Vinckboons. Unlike the Mannerists, these closely connected painters in blue worked in distinctive styles – no one will confuse Lastman’s history paintings with Vinckboons’s small histories depicted as part of the Southern Netherlandish landscapes. Moreover, the close thematic connections between artists working in different styles indicate that the choice of subject matter was made independent of the choice of style and location of residence. For example, Jan and Jacob Pynas had little overlap in their thematic choices despite being next of kin, and several Amsterdam painters, including Frans Badens ii and Pieter Isaacsz., had fewer connections with their city cohorts than with the Mannerists in other cities.

The most popular subjects in this period (fig. 5) include a great variety of mythological and Old Testament scenes favoured by the Mannerists. New subjects popularised by innovators like Pieter Lastman, such as Philip baptises the Eunuch, are also present in this network, albeit in the periphery. By far the most painted subject by both Mannerists and innovators was the Baptism of Christ. Fig. 5 also shows that the colour-coded subject clusters, or modules, do not fit in with the traditional categories such as mythology or the Old and New Testament. Instead, we see a mixture of all three in a single cluster. These clusters relate more to the visual resemblance of the scenes and the skills required. For instance, the Destruction of mankind (Old Testament) and the Marriage of Peleus and Thetis (mythology), both of which feature large groups of nudes in an outdoor setting, are to be found within the same cluster (marked in grey). This modular network structure heralds a transition from a conventional, small, and exclusive art world, in which artists were dependent on modest demand from the church or private patronage, to a commercial art market where paintings were sold as standardised commodities to a large group of ordinary citizens.27

Fig. 6 displays a simplified network of artists between 1610 and 1650, which, when measured by the number of painters, tripled in size compared to the previous period. Previous scholarship has amply demonstrated the growth in the population of painters in this period.28 However, scholars have not yet shown the structure of connections among these painters. Fig. 6 and tab. 1 illustrate that the number of links among artists increases almost sixfold (from 80 to 462), while the number of painters triples (from 31 to 97). In other words, this network’s interconnectivity (as measured by the average links per painter) is almost twice as large as that of other time periods. Next to the well-known fact that more artists picked from a larger pool of subject matters, fig. 6 suggests that each artist covered more variety of subjects in their own oeuvre instead of specialising in selective subsets. This observation challenges the general impression of deepening specialisation among seventeenth-century Dutch artists.29 Network analysis suggests that in fact the opposite took place at the subject level, adding weight to a new interpretation of specialisation that I have proposed elsewhere.30

Furthermore, the network of 1610-1650 is the least centralised among the three periods, meaning that it does not have dominant individuals and that connections are spread more evenly among artists. Fig. 6 also shows that Rembrandt, supposedly the greatest master of the century, did not enjoy the same central position as Cornelis van Haarlem had in the previous period. Other individuals with a similar level of connections to Rembrandt include Leonard Bramer (known for his numerous designs) in Delft, Jacob de Wet in Haarlem, and Cornelis van Poelenburch in Utrecht. It means that popular subject matters were disseminated among artists working in different market segments and were diffused beyond the city walls. The question remains, however, whether these observed changes in the network are not merely artefacts of a larger network that we might find in any network of the same size. We can check our observations by comparing them to randomly generated networks of the same size (Appendix ii). The simulation results suggest that the network of 1610-1650 is in fact much denser and less centralised than typical, random networks of the same size. This means that the growth of the population of painters alone cannot fully explain the large number of connections relatively evenly spread among painters – clearly, a shift had taken place in how artists picked their subject matters.

What had brought about the structural change in the network of iconography between 1610 and 1650? I would argue that the rapid expansion of the art market forced painters out of their old career path, bringing about a shift in marketing strategy. It was their adaptation to this ‘new reality’ that eventually shaped the network. Of course, none of these reasonings is entirely new, as socio-economic art historians have shown that the art market developed into a complex and competitive arena in which many painters and art dealers vied for an equally large number of clients.31 A relatively free market facilitated the flow of information across cities and smoothed the spread of subject matter throughout market segments. All of these market forces account for the profusion of connections in fig. 6 and tab. 1. Jan Steen, for instance, who is best known for his chaotic domestic genre scenes, occasionally also painted history scenes, drawing thematic links to various masters outside his city of residence. Jan Steen and his clients had obviously kept a close eye on the market.32

Economic art historians have suggested that painters were under market pressure to innovate and consequently introduced myriad new styles, techniques, and subject matters in this period.33 However, the innovative burst in subject matters between 1610 and 1630 faded away decades before the market contracted (see fig. 1).34 The art market seems to have continued to grow for another two decades despite the fact that the invention of new subject matters lagged behind.35 This observation calls into question Montias’s notion of ‘product and process innovations’. Although Montias elaborated on various process innovations as ways of increasing painting productivity and reducing costs, he remained vague on innovations that generated ‘new products’, nor did he sufficiently explain how key features in paintings such as subject matter, painting technique, and quality contributed to the introduction of these ‘new products’ in the art market. This is precisely where a network approach can offer new insights.

The colour-coded clusters in fig. 6 suggest that artists’ choices of subject matter, style, and quality were not connected. The Utrecht Caravaggist Gerard Honthorst had more subjects in common with Rembrandt than with his fellow Caravaggisti Dirck Baburen and Hendrick Bloemaert. Likewise, many painters in Rembrandt’s circle, including Gerbrand van Eeckhout and Salomon Koninck, fall into a different group than the great master. The close thematic ties between artists working with different styles and qualities in fig. 6 suggest that the demand for subject matter trumped stylistic preference. It was likely the combination of subject matter, style, and quality that defined a new product in the art market, even when none of them was novel. These multiple variables generated a myriad of combinations that kept the public interested in its diversity for decades, even after the thematic inventions dwindled by the 1630s (fig. 1). This preference for diversity encouraged an interconnected yet decentralised diffusion of subject matters. For this reason, subject matters popularised by great masters were soon picked up by lesser painters, whose lower-quality works of the same subjects were still embraced by the market.36 On the other hand, choosing the same subject matter means putting oneself in direct comparison to other masters. It explains why only Rembrandt’s direct rivals and best pupils – Jan Lievens, Govert Flinck, and Ferdinand Bol – could afford to share many subject matters with Rembrandt (marked in red in fig. 6), while other painters who worked in a similar style to Rembrandt’s, such as Salomon Koninck, avoided direct comparison by choosing different subjects (marked in blue).

In the network of subjects (fig. 7), mythological scenes – which dominated the subject network of the previous period – give way to the more conventional, even old-fashioned stories of the Old and New Testament, such as the Adoration of the shepherds and the Raising of Lazarus. Meanwhile, many new subjects formed their own cluster (marked yellow in fig. 7), including the Dismissal of Hagar and the Sacrifice of Manoah, which had been introduced by early innovators like Lastman and were subsequently popularised by successors like Rembrandt. However, except for a few outliers, most innovations in subject matters failed to gain momentum in the market, and painters still had to respect existing visual forms and traditional subjects if they were to appeal to buyers.

Perhaps surprisingly, the centre of the network between 1650 and 1700 is occupied by many painters known as fijnschilders – Willem van Mieris in Leiden, Adriaen van der Werff in Rotterdam, and Casper Netscher in The Hague – together with so-called Dutch Classicists like Gerard de Lairesse and Gerard Hoet (fig. 8). Even so, they belong to different clusters (blue and red, respectively), which suggests that a relatively small number of shared subject matters could be executed both as small, refined paintings and as larger, relatively rough tableaus. The prominent position of fijnschilders and Dutch Classicists in the network of iconography thus attests to the bifurcation of history paintings during the last decades of the seventeenth century. On the one hand, Classicist painters like Lairesse decorated palaces, public buildings, and mansions of the wealthy burghers, while on the other hand fijnschilders meticulously crafted small history pieces to enrich connoisseurs’ collections. The popularity of both groups speaks directly to the new fashion favouring both alternative wall-hangings (ceiling and mural paintings) and highly-refined easel paintings for collectors, substantiating the ‘aristocratization’ (veradelijking) observed by art historians.37

Furthermore, the strong ties among fijnschilders signal a change in competition and marketing strategy. Painters of the same style and quality no longer avoided using similar subjects to their competitors. Rather, they embraced direct comparison to show off their skills, achieving the same effect through different techniques – what Angela Ho has called ‘creative repetition’.38 The patrons of the fijnschilders seem to have favoured this ‘creative repetition’ over more experimental subjects, which, in turn, steered the competition towards advanced painting techniques instead of inventing novel subject matters.

This ‘creative repetition’ is also evident in the subject network (fig. 9). The connections among the subjects depicted by painters after 1650 are more evenly distributed, and the network is less centralised – no dominant subjects can be found in this time period.39 It suggests that the painter often picked the same popular scenes instead of exploring new iconographies. Expanding repertoires would be too risky for fijnschilders, because, as Junko Aono has argued, affluent collectors valued these painters’ works when they inherited and updated the pictorial tradition of the old masters.40

In this period, the subject matters picked for paintings were increasingly drawn from mythology and the Old Testament. Painters seem to have abandoned the new inventions from the previous generation and resorted to the more traditional scenes like Susana and the elders and Lot and his daughters, constraining themselves to an even smaller collection of subjects. Fijnschilders and their patrons favoured the artistic execution over the subject matter, leaving little room and need for thematic innovation. The popularity of a limited number of subjects and the virtual absence of others in the network of subject matter attests to Eric Jan Sluijter’s observation that the period saw a highly selective and restricted repertoire.41 As Marten Jan Bok has put it, ‘creative freedom was an ideal rather than a reality’.42

The previous section broadly outlined how the structural change in the network of iconography reveals shifts in artistic choice, competition, and trending subject matters. This section takes a closer look at a group of well-researched painters in Amsterdam, using them as a case study to analyse how the network of iconography relates to the artistic ‘influence’ and connections identified in Eric Jan Sluijter’s Rembrandt’s Rivals.43 In this seminal work, Sluijter examined the artistic competition as a conscious rivalry among artists, who both imitated and emulated each other in composition, manner (handeling), motif, and subject matter. In his analysis he draws connections between different painters, which implicitly weaves a network of artistic rivalry through influential relations. Fig. 10 visualises this rivalry network in Rembrandt’s Rivals.

The node size represents a master’s level of influence, measured by how many other artists picked up or ‘borrowed’ ideas from him. The thickness of the link between nodes represents the strength of such artistic ‘borrowing’, ranging from acquiring the source’s style (such as the thick line between Rembrandt, Flinck, and Bol), to employing his compositions or motifs, and to there being merely the faintest of resemblances between the works (i.e., a thin line). The colour-coded clusters (modularity classes) in fig. 10 corroborate the grouping by style and manner as understood by art historians. But how does the iconographic network of artists relate to the rivalry network, and how do these two networks compare to the social network? To answer these questions, I have filtered out the artists featured in Sluijter’s study and created an iconographic network among Rembrandt’s rivals (fig. 11).

Comparing the networks based on artistic rivalry and iconography, it is clear that the great masters who had a powerful stylistic impact (like Rembrandt or Lastman) also occupy prominent positions in the iconographic network. Nonetheless, the thematic network cannot distinguish innovators or trendsetters from their followers, as the latter – who kept a close eye on the innovators – could take on the same subjects. For this reason, the followers occupy more critical positions in the thematic network than they could in the rivalry network. Bartholomeus Breenbergh was one such astute follower, who stands out in fig. 11 for his numerous connections. Breenbergh, known for his Italianate landscapes with small figures, often inserted elements of Rembrandt’s and Jacob Backer’s works into his renowned landscapes, such as in the Preaching of St. John the Baptist.44 In this way, Breenbergh relied on his ‘unique selling point’ while staying in tune with what was popular at the time. The difference in style and manner freed him from direct competition with the great masters. Duplications in subjects would have made Breenbergh’s works more attractive to his clients.

Sluijter suggests that many painters active in Amsterdam between 1630 and 1650 were oblivious to the new painting style introduced by Rembrandt and instead clung to well-established traditions and painting styles. The network of iconography reveals that it is not that they did not react to Rembrandt’s innovations, but that they did so in a new way: painters rode the tide of popular subject matters introduced by Rembrandt and other innovators while sticking to their own style. Conventional painters like Adriaen van Nieulandt and David Colijns seem to have actively participated in the competition by picking up popular subjects, even though they did not adopt Rembrandt’s new painting style.45 In other words, even though artists produced topical and novel subject matters, they often did so in conventional styles. This observation underlines the fact that competition occurred not only in style, but also in subject matter. This was particularly suitable for a large and competitive art market like Amsterdam, which was capable of both accommodating and appreciating various styles depicting an identical (and often popular) story.

Fig. 12 visualises the painters who lived in Rembrandt’s neighbourhood in squares, in order to see if spatial proximity was reflected in closer bonds in style or subject matter. However, both networks show little trace of artistic and thematic transmission facilitated by proximity. Perhaps counter-intuitively, the style and thematic connections between Rembrandt and Moyaert/Breenbergh, who never resided in the same neighbourhood, are much stronger than those between Rembrandt and Van Nieulandt, even though the latter lived across the street from Rembrandt for over a decade.46 The choice of style and subjects was likely rooted more in one’s ability, ambition, and marketing strategy than in location. It further suggests that artistic innovations flowed freely throughout Amsterdam and beyond, making residence and workshop location an insignificant factor for accessing information.

What about the social network? Many scholars have employed social networks to infer the transmission of ideas, yet the social network of Rembrandt’s rivals (fig. 13) suggests no further diffusion of subject matters beyond those taught by masters to their pupils. Painters working in similar styles (such as Salomon Koninck and Rembrandt) or sharing many subject matters (Moyaert and Breenbergh) had somewhat distant social relations.47 Therefore, social ties alone can hardly explain knowledge dissemination at scale.

What stories were favoured by painters in Amsterdam? The network of subject matter (fig. 14) shows that Amsterdammers endorsed more unconventional subject matters, many of which were introduced and popularised by Pieter Lastman in the early seventeenth century. The Dismissal of Hagar and the Sacrifice of Manoah stand out as the most popular scenes in fig. 14. These subjects had attracted painters ranging from Rembrandt to minor masters, such as Johannes Urselincx, who presented their own interpretation of the story in various styles and qualities.48

Lastman’s Sacrifice of Manoah seems to have triggered a wave of sacrifice scenes, echoing the popularity of such stories in the theatre at the time.49 Numerous sacrifice scenes were depicted based on both heathen and sacred stories to fit into the sacrifice theme. Several painters picked up obscure stories like the Sacrifice of Polyxena, the Sacrifice of Iphigenia, and the Sacrifice of Jephthah’s daughter, all of which share the iconography of a female victim confronting her fate at an altar.50 However, very few examples of Jephthah’s daughter are mentioned in inventories in the Montias database, and Polyxena or Iphigenia are missing altogether from the Getty Provenance Index – although there are many entries listed as a sacrifice scene without explicit references to a single story.51 Perhaps contemporaries were more concerned with the act of sacrifice than the individual girl in these instances: it was the dramatic, even brutal act of sacrifice that stood out in their eyes.

The disparity between the networks of artistic rivalry and that of iconography further substantiates the earlier observation that for most painters, the choice of subject and style were unrelated. They fed the market with paintings depicting popular subjects in various styles, and perhaps used stylistic variance as a deliberate marketing strategy to differentiate themselves on the market, catering to different demands. This strategic differentiation is evident from the central location of several Hagar scenes of various qualities (fig. 15). After Lastman’s work of 1612, the Dismissal of Hagar (fig. 16) seems to have swept the Dutch Republic. The production of this Hagar scene peaked between 1640 to 1660, even though demand for Old Testament subjects had declined after 1640 (fig. 1).52 Unlike the sacrifice scenes, the paintings of Hagar were well recognised by the market, as they regularly appeared in contemporary inventories. In Amsterdam, for instance, some forty-one Hagar-themed paintings were either inventoried or sold in auctions and shipped abroad.53 Another sixteen hung in houses in Haarlem, and seven more in Dordrecht.54

What made the Dismissal of Hagar so popular? Art historians have argued that the story is an expression of human emotions – jealousy, pride, loss, sorrow, and fear – that are all reflected in the twists and turns of the dismissal scene.55 Jan Pynas even included a peacock (fig. 17), a traditional symbol of pride and arrogance in Dutch art and literature, to signify the cause of the banishment. ‘So pride comes before the fall’, as Jacob Cats wrote in his Lamentation of Hagar. Christine Sellin has argued for the didactic value of the Hagar story, as it was used in moralist pamphlets, literature, and sermons.56 The congenial social environment and the catchy and emotive content of the story of Hagar could have created a general interest in the visual representation of the Hagar narrative from the market. That said, most of the literature and sermons that Sellin assembled in support of her argument did not appear until the early eighteenth century, long after the wave of Hagar paintings had subsided (fig. 15b). It seems that aspects of Dutch society other than the art market did not dictate artists’ choice of subject matter in this case. Nor could it explain why only the expulsion scene (Gen 21:14) gained momentum in visual arts, not the scenes discussed in moralistic writings, such as the her running away (Gen 16:7-12).57 While it is true that the Dismissal of Hagar would not have been popular without a supportive social circle, these external forces alone cannot fully account for the Hagar phenomenon.

The missing factor, I would argue, lies in the artistic impulses. The vogue of Hagar paintings epitomised the instance when market interests and artistic pursuit crossed paths. The story caught the attention of influential masters such as Lastman and Rembrandt, who strove to depict the passions in their work.58 The expulsion episode might have appealed to these artists for its suitability to study the relationship between strong emotions, subtle movement, and dramatic narrative, all of which fit their artistic pursuit and ambition.59 Rembrandt is said to have assigned this subject to his pupils as a study, as most extant drawings of the expulsion scene come from Rembrandt’s circle.60 The diffusion of the Dismissal of Hagar can be seen as an interplay between artistic pursuits and market interests: the artistic pursuit in depicting a complicated and emotionally charged scene was appreciated by the market and ignited the public’s enthusiasm. The benefit of this fortuitous meeting of artistic and market interests spilt over to the lower end of the market and eventually allowed the Dismissal of Hagar to become a breakaway success.

This essay has revisited Montias’s ‘process and product innovations’ and re-evaluated the artistic connections among Rembrandt’s rivals by developing a novel network of iconography, which connects artists, for the first time, not through social relations but through shared subject matters. This new art-historical network examines history paintings in ways that traditional methods cannot and illustrates how scores of painters in the seventeenth century were connected artistically – and how such connections were translated into iconographical inventions. It further demonstrates that the choices of style and subject matter were often unrelated, and that thematic connections between painters often had little to do with their social relations or physical proximity. These findings suggest that painters did not need to know or see each other’s work to pick up a certain subject. Conversely, the choice of subject matter was often influenced by market forces: artists at all levels produced paintings of the same popular subjects, though executed in distinctive styles and qualities to increase diversity and maintain a competitive edge. As such, these findings help to understand what defines a new product in Montias’s ‘product and process innovation’. Nevertheless, market forces were not omnipotent. It also took painters’ artistic pursuits in conjunction with market interests to allow a thematic innovation to sweep across the board, as did the Dismissal of Hagar. This approach also sheds new light on the structural shifts in the seventeenth-century art market and the ways in which painters adapted to changing patterns of demand and style.

Finally, it is worth stressing the exploratory nature of this study, as it raises more questions than it can answer. Questions regarding a possible correlation between the shift in market competition and the political, religious, and economic context of Dutch society, as well as the impact of foreign connections on networks of iconography, still remain unexplored. Due to the limitations of the data source, the findings presented in this article are necessarily limited in scope. Therefore, the main contribution of this study lies more in its methodological exploration than in its discoveries. It shows that the application of computational methodologies, if well thought out, can provide new interpretations through linking various datasets. Hopefully, the current digitisation process that promotes Linked Open Data will provide researchers with new information regarding the Dutch art market. Until then, this essay has hopefully demonstrated how (digital) art historians might approach the bridging of digital and conventional methodologies.