Re-Views: Field Editors’ Reflections
caa.reviews

With “Reflections on Digital Art History,” caa.reviews inaugurates a new field of coverage, since our future is now. In fact, immediately prior to drafting these remarks, I noticed a headline on Hyperallergic.com asking, “Can an algorithm determine art history’s most creative paintings?” I was only curious enough to skim a paragraph or two, yet surely many of us sympathize with the convergence it represents. On the one hand, popular imagination and political rhetoric have increasingly figured the humanities as superfluous to the needs of civilization. On the other, STEM (science, technology, engineering, and mathematics) research has given us tremendous computational powers and capacities for information storage and dissemination.

Maybe we are late to the game. As Pamela Fletcher, caa.reviews founding Field Editor for the Digital Humanities and Art History, writes in her following reflections: “It is the rare art historian who has not already had her or his teaching, research, and publishing significantly impacted by the technological changes referred to in shorthand as ‘the digital.’” Our challenge is not simply to react, but to embrace and innovate. The digital humanities is an emerging field that promises to merge technology with interpretative creativity, to create entirely new types of scholarly perspectives and methods for visualizing data, and to communicate these discoveries far and wide. Books were wonderful, but they belonged to a flat earth.

In her introductory essay, Fletcher identifies two existing categories of projects that are already showing potential: archival modeling and visual data analysis. The former seeks to assemble tremendous detail in one central location, such as every document directly tied to Dante Gabriel Rossetti’s art, while the latter seeks to make complicated time and space relationships comprehensible, such as the evolving physical plant of Auschwitz-Birkenau during World War II. The scholarship is already inspiring, since it is these pioneers who are setting the course we will all soon travel.

David Raskin, Editor-in-Chief, caa.reviews

REFLECTIONS ON DIGITAL ART HISTORY

It is the rare art historian who has not already had her or his teaching, research, and publishing significantly impacted by the technological changes referred to in shorthand as “the digital.” Slide libraries are obsolete, replaced by digital images and databases; images of variable quality and accompanying information (metadata) flood the internet; journals and presses publish in digital formats; and libraries, archives, and museums are digitizing the objects and texts in their collections. But none of this is primarily what we mean when we talk about “digital humanities” or even “digital art history.”

The digital humanities is by now an established scholarly designation and field, with academic journals, conferences, graduate programs, and a dedicated office at the National Endowment for the Humanities. The term, which entered popular usage in 2004 as the title of an anthology edited by Susan Schreibman, Ray Siemens, and John Unsworth, is used to refer to the massive changes in humanities scholarship made possible by high-speed networked computing, digitization and big data, and increasingly sophisticated algorithms.¹ The “What is Digital Humanities” essay has become a subgenre of its own, numbering countless blog posts, essays, and books.²

The term—and perhaps the practice of—“digital art history” is less firmly established, though it has recently gained considerable traction, as evidenced by recent conferences at the Institute of Fine Arts (“Digital Art History” in 2012) and the Center for Advanced Study in the Visual Arts at the National Gallery of Art (“New Projects in Digital Art History” in 2014), a special issue of Visual Resources in 2013 entitled “Digital Art History,” the recent summer workshops on digital art history sponsored by the Samuel H. Kress Foundation and the Getty Foundation, and the new International Journal for Digital Art History (www.dah-journal.org), to name only some of the most prominent examples.³ Recent reports and articles by William Vaughan, Diane Zorich, Johanna Drucker, Anne Helmreich and Murtha Baca, Michelle Millar Fisher and Anne Swartz, and Paul B. Jaskot and Anne Collins Goodyear begin to define (and debate) precisely what digital art history is and what its potential might be.⁴ These essays offer very useful overviews of the discipline’s engagement with technology and raise important questions, including the challenges that new forms of knowledge production raise for teaching, publication, and evaluation; the perception of widespread skepticism about the transformative intellectual potential of digital and computational methods; and art historians’ responsibility to work with technologists to build tools better suited to engaging humanistic questions and values. Overall, the authors share the sense that while the digitization of art-historical materials has made real progress, there are fewer well-known examples of an interpretative digital art history. In a recent review of the anthology Debates in the Digital Humanities, Jaskot argues, “The question is not what art history can do with the digital; the question is what are the important art historical questions that can be addressed with the help of digital tools?” and he calls for a digital art history that “[puts] the intellectual problem (rather than a method) at the center of the discussion.”⁵

One of the goals of the new field editorship in “Digital Humanities and Art History” for caa.reviews is to help answer that call, bringing more visibility to digital art history projects and creating space for discussion of both their art-historical and technological achievements. In this review essay, then, I will focus on the intellectual potential of the intersection between new technologies and the discipline of art history, rather than taking up the many important institutional, professional, and ethical issues raised by these new methods and their implementation. In the first section, I discuss digital projects that aim at increased access, making resources (images, archives, catalogues) available digitally. In the second section, I turn to projects that use computational methods to analyze and interpret art-historical materials. This is not, of course, a complete list of such projects; nor is it a comprehensive overview of all potential digital and computational methods. Instead the projects have been chosen as examples, meant to suggest possibilities and ideas both to those interested in adopting digital methods and to those simply curious about the parameters of this emerging field.

I. Digitizing Art History

Some of the first digital projects of scholarly interest to art historians took the form of repositories of digital texts and images, catalogued, organized, and made available online. Three scholarly projects give some sense of the range of possibilities.⁶ Jerome McGann’s Rossetti Archive brings texts, images, and significant editorial content together, aiming “to include high-quality digital images of every surviving documentary state of D[ante] G[abriel] Rossetti’s works: all the manuscripts, proofs, and original editions, as well as the drawings, paintings, and designs of various kinds.”⁷ These materials are elucidated by editorial content, including catalogue entries and bibliographies; a platform for allowing scholars to share their own images and scholarship is currently in development. Closer to Van Eyck: Rediscovering the Ghent Altarpiece, a project led by Ron Spronk, takes a different tack, focusing on a single object rather than the totality of an artist’s career. During an examination of the painting to determine its conservation needs, thousands of high-resolution images of the altarpiece were taken in extreme close-up using macrophotography, infrared reflectography, and x-radiography, then reassembled using the program MATLAB.⁸ The website interface allows viewers to zoom in on images to an astonishing level of detail and to compare different parts of the altarpiece and different types of photographic imaging. In contrast, Mapping Gothic France, a project led by Stephen Murray and Andrew Tallon, aims to bring together documentary materials about a class of structures—Gothic cathedrals—in a way that allows for multiple visualizations and narratives about the formation of the Gothic and the formation of France in the twelfth and thirteenth centuries.⁹

This type of work is important intellectually for several reasons. First, as Tom Scheinfeldt and William Pannapacker have argued, building tools and creating reference materials—however undervalued they may currently be as scholarly work—have always been the necessary foundations upon which knowledge is built.¹⁰ Art historians know this well: indeed, catalogues raisonnés, museum catalogues, and provenance archives are critically important for art-historical research and interpretation, and much current work is built on the foundations provided by the generations of scholars who spent, in some cases, entire careers compiling these resources. As research materials continue to be digitized and organized in new ways, it is important to rethink how to value and reward the work of building digital archives and catalogues. Second, many of these tools and databases allow not just access to data, but the ability to compare elements—images, floor plans, and details—easily and quickly, enabling the user to make new connections.

Such projects are intellectual labor, however, at an even deeper level. As Drucker argues, the digitizing, organizing, and devising of metadata are always acts of interpretation, the product of a set of decisions that “carry interpretative inflection: they are not neutral or value-free, and each privileges one aspect of a digital artifact at the expense of others.”¹¹ The choices made at the point of digitizing an image and creating a database—from the lighting conditions and angle at which an object is photographed to the categories of metadata that are used—create a canon of sorts, not only of which objects are (and are not) digitized but one that is built into the very material and conceptual ways in which they are represented.

II. Computation and Art History

A second category of digital art history projects brings computational methods to bear on art-historical questions. This form of scholarship is premised on two things. First, the increasing availability of “big data”—or at least “bigger” data: the raw material of images, texts, and metadata available in digital formats. The second is the increasing availability of computational analytic tools—many originally developed for scientific or commercial use. Elijah Meeks, a digital humanities specialist, summarized the analytic tools most useful to humanists, identifying “three pillars to DH research: Text Analysis, Spatial Analysis and Network Analysis,” and then adding, “I continue to have a sneaking suspicion that Image Analysis is something else that sits with the aforementioned three.”¹²

Text analysis has dominated literary digital humanities, though it is used comparatively less often in fields such as history and art history.¹³ Computers can sort through vast amounts of text rapidly, finding patterns of letters, phrases, and even identifying topics in chunks of texts. In machine learning, computers find similarities across hundreds or thousands of texts, sometimes confirming preexisting humanly devised categories such as author or genre, and sometimes making visible new categories into which texts might be grouped. Literary scholars Franco Moretti and Matthew Jockers have published a provocative series of papers and books reflecting on the potential of such work—what Jockers calls “macroanalysis”—arguing that the literary field should be “approached not simply as an examination of seminal works but as an examination of an aggregated ecosystem or ‘economy’ of texts.”¹⁴ Their research both raises significant methodological questions about scale and the relationship between “distant reading” (Moretti’s term) and close reading, and suggests some potential methods and research questions that might be productively taken up within other disciplines. For instance, Jockers provides a fascinating example of how available metadata might yield rich insights, analyzing titles and identifying attributes of 758 works of Irish American literature over 250 years. The resulting visualizations track variations in the prevalence of novels by women versus men, in the eastern versus western United States, and provides information about title lengths and the frequency of title subjects over time.¹⁵ This is fairly simple computational work, but the picture it paints of the overall field of production is rich and surprising. What results might be found if similar analysis was conducted of titles and subjects of works of art in auctions, exhibitions, and collections?

Spatial analysis—a term I use to include both geographical mapping and various forms of visual reconstruction—is another area that has seen significant growth in recent years as part of an increased attentiveness to “space” as an object of study and analysis.¹⁶ Historians and cultural geographers have made tremendous use of Geographic Information Systems (GIS), giving rise to an entire field of “Historical GIS.”¹⁷ Art historians have begun using GIS and other mapping tools to track the movement of artists and artworks across geographical territory. Jodi Cranston’s interactive website, Mapping Titian, allows users to trace the movement of Titian’s canvases over time, providing the tools for an enriched history of reception by making visible the historical proximities of specific pictures and people; while Michele Greet’s Transatlantic Encounters: Latin American Artists in Interwar Paris uses maps to illuminate the social networks of Latin American artists in Paris.¹⁸ Mapping the movement of people, artworks, and art institutions across geographic space reveals patterns and intersections that otherwise remain invisible; in the words of historian Edward L. Ayers, in dynamic maps, “The patterns, intricate and shifting, are too complex to easily explain in words or even numbers. We can see more in maps than we can easily say.”¹⁹ The ability of mapping programs to manage and visualize large data sets also allows for an expansion of scale; the directors of ARTL@S, an ambitious project which aims to create a mapping platform that allows scholars to share data, produce maps, and publish scholarship, are inspired by the belief that the combination of quantitative analysis and a geographic approach can help move art-historical questions beyond nationalistic accounts.²⁰

Geographic visualizations also raise challenging questions about the relationships between space, time, and experience. Most available mapping tools represent change over time through animations of snapshots of particular sets of mapped points. Such visualizations can be tremendously useful, as in Jaskot, Anne Kelly Knowles, Chester Harvey, and Benjamin Perry Blackshear’s work on mapping the built environment at Auschwitz.²¹ Adding time signatures to their geographical data and creating an animation revealed that Auschwitz-Birkenau was under nearly continuous construction between May 1943 and February 1944, suggesting the “camp was visually confusing, quite a different environment than the regimented, rational, static image of the camp that has become so familiar to us.”²² This shift in viewpoint raises radically new questions about the lived experiences of Auschwitz for both inmates and SS personnel, opening new avenues of historical exploration and interpretation.

But there is also considerable information about temporality that such animations cannot convey. For example, as I worked to build a digital map of art galleries in Victorian London, I quickly realized that at any specific moment in time some galleries would have seemed familiarly old to the viewer who frequented Bond Street, while others would be newly established spaces.²³ This type of knowledge would be an obvious and potentially quite meaningful aspect of place to viewers, but it is invisible on a static representation of time on a map. Might we work with computer scientists to build better tools, conceptualizing a map differently in order to convey the kind of temporality that lives within the instantaneous?

Moreover, time itself, like space, is culturally variable. Drucker has beautifully articulated the challenge of representing humanistic understandings of time and space within the positivist templates of satellite maps, time measured by clocks, and digital databases:

Both space and time are constructs, not givens. As constructs they come into being in a codependent relation with their discursive or experiential production. If I am anxious, spatial and temporal dimensions are distinctly different than when I am not. When the world was bounded by the Mediterranean it was a different world than the one seen from space. These are not different versions of the same thing.²⁴

Art historians—familiar with other conventions for representing space and time, such as medieval maps in which time and space are codependent or maps of Roman roads that represent the distance between cities and not their relative position—have much to contribute to conversations about the interrelation of space, time, and experience, and how these qualities might be represented computationally.

Reconstructions—of exhibitions, of lost architectural monuments, and even of entire cities—are also rich areas for scholarly exploration. Art and architectural historians have created and/or are working on recreations of the ancient Egyptian temple complex at Karnak, the Roman Forum, the city of Venice, and medieval monasteries.²⁵ Such virtual reconstructions offer tremendous potential for asking questions about the lived experience of spaces, including usage, sight lines, and acoustics.²⁶ Diane Favro and Christopher Johanson’s article on funerary processions in the Roman Forum is an outstanding example of scholarly innovation in this area. After having built a virtual model of the Forum, they used it to visualize three different Roman funeral processions, aiming to understand the visual and aural experience of those processions, but also testing and evaluating alternative reconstructions of the historical data.²⁷

Researchers involved in such projects are also cautious about the fantasy of perfect historical accuracy. As projects increase in sophistication, they can seem to provide a total immersion in an alternative world, along the lines of Renaissance Italy as recreated in extraordinary detail for the video game Assassin’s Creed II. How do scholars represent relative degrees of uncertainty in their recreations? ²⁸ How do they choose the specific historical moment of re-creation, and what do those choices say about interpretive assumptions? In a recent team-taught course on Digital Humanities at Bowdoin College, one student project aimed to reconstruct the famous 1913 Armory Show in New York, but ended up stymied by the lack of data about the specific hanging of works on the walls.²⁹ In class, we began to discuss what precisely we wanted to reproduce in such projects, and posed the possibility that it might be at least as useful to our historical understanding (and potentially more fun) to confront the problem faced by the organizers of hanging the hundreds of works in the space in a short period of time, rather than re-creating the final results. Thinking through the limits and possibilities of virtual reconstructions can prompt us, then, to ask which historical moments and experiences we identify as the proper subject of analysis, and what those choices suggest about our methodological and interpretive assumptions.

Network analysis, the “third pillar” of digital humanities, is a mathematical modeling of relationships between entities.³⁰ While it can be disconcerting for the visually oriented to realize that network visualizations are not necessarily based in geography or spatial relationships—the same data can be visualized with nodes in different places, as the point is the relationship between them, not their relative position—the mathematical power of network analysis is vast. One obviously useful application is identifying the social networks of artists, such as Michelle Moravec’s work on “Visualizing Schneemann” or the graph produced as part of the Museum of Modern Art’s show Inventing Abstraction, 1910–1925.³¹ Helmreich’s network analysis of the Goupil Gallery’s stock books extends this line of inquiry to objects, tracking patterns of circulation through the nineteenth-century international art market of dealers and collectors.³² As scholars become more familiar with this method, harnessing the statistical power of network analysis in addition to using the visualization capacity offers rich potential.

Image analysis is a very active topic in computer science and in the race to build better search engines and surveillance technology, but its scholarly applications to the study of art history are still in their infancy. John Resig has done tremendous work bridging this gap, both in his presentations on computer vision and art history at THATCamps held at the College Art Association conferences in 2013, 2014, and 2015, and in his work on the website Ukiyo-e Search.³³ The site allows viewers to upload their own photographs of Japanese woodblock prints and to search for similar images among over 200,000 prints from museum collections and other databases, allowing for both identification and comparison. Resig is currently exploring how this technology could also aid in the study of the vast photographic records of paintings, drawings, and prints held in various archives.³⁴ Other computer science applications of interest to art historians include analysis of brushwork and gesture in the paintings, respectively, of Vincent van Gogh and Jackson Pollock; diagramming the perspectival systems of individual paintings in order to think about when and why artists deviated from these mathematical models; and using machine learning to explore the iconography of medieval book illuminations.³⁵ Facilitating even more conversations between computer scientists and art historians is a necessary next step, as defining questions of shared scholarly interest will expand research possibilities in both directions.

So what larger connections do these examples suggest between available computational methods and art-historical questions? I think there are at least three potential areas for fruitful inquiry and methodological reflection. First, the techniques of computer vision and machine learning offer intriguing possibilities for the study of attribution and materiality. While attribution may sound like an “old-fashioned” question, it remains foundational work for the discipline. Computer-aided study of how specific objects were created, combined with the ability to compare and sort them in relation to other ones, could aid in seeing larger patterns in workshops or changing uses of different materials over time in ways that might enrich current discussion of materiality, process, and artistic influence. Second, the ability of computers to analyze large data alters the temporal and geographic scales at which art-historical inquiry can take place, shifting attention to larger and longer patterns in practice and reception. For some scholars these larger trends will be their primary subject, but for others they will be an enriched context, the “signal” against which the “noise” of the exceptional singular example becomes more meaningful. In either case, the shifts in possibilities between distant and close reading should prompt methodological reflection on the meaning of “context” and the specific contours of the relationships between individual people and objects, as well as larger historical patterns. The collaborative nature of digital art history projects can also shift the scales at which we work; as the examples I have shown suggest, this kind of research generally requires partnering with other scholars whose different expertise and perspectives can enlarge both the scope of work possible and the interpretative breadth. Finally, the move to digital space and visuality poses questions that art historians may be uniquely suited to answer. Many colleagues in digital humanities worry about the implications of a shift from text to image, as a text-based literary scholarship moves into a digital space fundamentally mediated by screens and vision. In the face of this fear of the power of the image, art historians’ long history of interpreting and analyzing the rhetoric of the visual can be of real use, particularly in the use of digital means to reconstruct and represent lost objects, buildings, and spaces, and in so doing interrogate the fantasies the digital offers of “perfect” reconstruction and virtual access to the absent.

III. Conclusion

Digital art history is not—or does not have to be—a separate track within the discipline, available only to the technologically inclined or gifted. Building new tools and contributing to the creation of technology that is attuned to humanistic questions and values is one important task. But it is also the case that there may not always be a correlation between the most advanced computational techniques and the most compelling art-historical interpretations. Many comparatively simple tools for mapping or network analysis can yield insights that transform a scholar’s understanding of a subject, organizing and visualizing even relatively small amounts of data in ways that make patterns and relationships more visible. The expanded data sets that digitization makes possible will allow some scholars to define a larger context for their objects of study, while keeping their focus on the close readings of individual artists and objects. In the end, digital and computational techniques can be significant additions to the art historian’s methodological tool kit: machines for thinking with, rather than replacements for thinking. As historian Richard White eloquently explains: “visualization and spatial history are not about producing illustrations or maps to communicate things that you have discovered by other means. It is a means of doing research; it generates questions that might otherwise go unasked, it reveals historical relations that might otherwise go unnoticed, and it undermines, or substantiates, stories upon which we build our own versions of the past” (emphasis in original).³⁶ The entire field of digital art history—and probably even digital humanities—is at the very beginning of a long research experiment, and the very forms of questions that it will pose—let alone the answers it will help us to generate—are unknown territory.

Defining digital art history and its relationship to the larger fields of digital humanities and art history is thus a collaborative work in progress. As this work is being done, it is worth asking how our own specific disciplinary history intersects with the digital and computational. Historians Scheinfeldt and Stephen Robertson have recently called for attending to the “disciplinary differences in digital humanities” ³⁷ and have begun outlining a genealogy of digital history that has its origins in oral history, folklore studies, and public history rather than the more familiar digital humanities origin narrative of humanities computing and text analysis.³⁸ What might digital art history’s alternative genealogy look like? Its origins might lie not in Father Roberto Busa’s use of IBM computers to index the work of Thomas Aquinas, but in the history of the reproductive technologies of photography and film, lantern slides and pixels. It might include the pioneering work of art historians like Jules Prown, who in the 1960s used computers to study John Singleton Copley’s patronage, and the work of the members of the Computers and the History of Art Group (CHArT)—self-identified as a group of “art and design historians who happened also to be computer enthusiasts”—who have been meeting and collaborating since 1985.³⁹ It would extend to artists’ experiments with technology from the first hand-held video recorders to the internet, and include the work of museums and cultural heritage groups in using digital technologies to give greater public access to works of art and architecture.⁴⁰ Telling this story might help in identifying the commitments, questions, and assumptions undergirding a digital art history, and situate it within a longer trajectory of art history rather than understanding it (only) as a foreign import from the scientific and commercial worlds. Technological change, as many are quick to warn, may well be inevitable, but the development of an intellectually generative digital art history is up to us.