Painting by Numbers

Data-Driven Histories of Nineteenth-Century Art

Diana Seave Greenwald

• Publisher: Princeton University Press
• Year: 2021
• Print ISBN: 9780691192451
• eBook ISBN: 9780691214948
• Topic: Art
• Subtopic: European Art

A pathbreaking history of art that uses digital research and economic tools to reveal enduring inequities in the formation of the art historical canon

Painting by Numbers presents a groundbreaking blend of art historical and social scientific methods to chart, for the first time, the sheer scale of nineteenth-century artistic production. With new quantitative evidence for more than five hundred thousand works of art, Diana Seave Greenwald provides fresh insights into the nineteenth century, and the extent to which art historians have focused on a limited—and potentially biased—sample of artwork from that time. She addresses long-standing questions about the effects of industrialization, gender, and empire on the art world, and she models more expansive approaches for studying art history in the age of the digital humanities.

Examining art in France, the United States, and the United Kingdom, Greenwald features datasets created from indices and exhibition catalogs that—to date—have been used primarily as finding aids. From this body of information, she reveals the importance of access to the countryside for painters showing images of nature at the Paris Salon, the ways in which time-consuming domestic responsibilities pushed women artists in the United States to work in lower-prestige genres, and how images of empire were largely absent from the walls of London's Royal Academy at the height of British imperial power. Ultimately, Greenwald considers how many works may have been excluded from art historical inquiry and shows how data can help reintegrate them into the history of art, even after such pieces have disappeared or faded into obscurity.

Upending traditional perspectives on the art historical canon, Painting by Numbers offers an innovative look at the nineteenth-century art world and its legacy.

CHAPTER 1

What Is a Data-Driven History of Art?

In the mid 1960s, the art historian Jules Prown was jeered. He was presenting new research at the annual meeting of the College Art Association, the principal professional art historical organization. Prown’s first slide—which showed an IBM punch card, then representative of cutting-edge computing technology—prompted some colleagues to boo.¹ In his essay “The Art Historian and the Computer,” Prown describes how he arrived at that presentation. He used a mainframe in the Yale University computer lab to examine the link between the socioeconomic backgrounds of American painter John Singleton Copley’s sitters and their preferences in portraiture. Recounting this experience, Prown writes:

At first consideration the art historian and the computer would seem to be eccentric companions. Art historians are concerned with qualitative discriminations that reveal themselves slowly, and at an unpredictable tempo, to the investigations of a trained mind and sensitive eye. The computer, on the other hand, deals with quantitative computations at an unvarying pace with incredible speed. Its monotonous, inflexible, unthinking efficiency sends a shudder down the spine of any self-respecting art historian.²

Yet as he discovered, these apparent enemies can work together.

Prown—famous for being a passionate advocate of formal analysis and close looking—described his impulse to use computers while preparing a monograph on Copley. Discussing the roughly 350 artworks that Copley created while he was working in North America, Prown writes that “certain questions c[a]me insistently to mind.” These were mostly about the social makeup of Copley’s clientele over time: “Were Anglicans wealthier than Congregationalists, and if so, did Copley paint more Anglicans as he himself prospered? . . . Did merchants order bigger pictures than ministers or vice versa? Which group ordered more pastels?” For Prown, “it seemed quite clear that these questions and many similar ones about the patterns of Copley’s patronage could be answered through a statistical analysis of his paintings and their subjects.”³ Statistical analyses did indeed indicate that certain kinds of patrons—defined by their occupations, political affiliations, and other social attributes—were more likely to purchase a work from Copley at different periods in the artist’s career. Sitters of certain professions were also more likely to purchase particular sizes of works. With the help of a computer and statistics, Prown was able to see trends in Copley’s career and oeuvre that were otherwise invisible. The quantitative view complemented the qualitative one.

At the end of his essay, Prown concluded: “On the basis of this experience . . . I am convinced that the computer can and will be used fruitfully for other studies in art history. The computer can be especially helpful for projects which require any kind of quantification.”⁴ Despite the fact that his colleagues dismissed this work at the College Art Association—after his presentation “a senior art historian berated [him] loudly in the aisle for [his] apostasy”—Prown made a prediction about the future of art historians and computers.⁵ He said that art historians would happily adopt computing technology as an automated retrieval system for images and information, but they would resist its more complex statistical uses.⁶ His prediction has generally proved true. This book, however, seeks to return to the complementary data-driven modes of inquiry that Prown first engaged with more than fifty years ago.

I trained as both an art historian and an economic historian. These two disciplines—like the art historian and the computer—may at first seem incompatible. The methodological foundations of art history are, as Prown describes, close looking and focused examination of select artists and works of art. The social sciences, including economics and economic history, are often quantitative and dependent on large datasets; the scale of social scientific evidence is orders of magnitude greater than the number of paintings art historians typically analyze. Furthermore, economic methods provide a zoomed-out view of history, where people, events, and historical change can be reduced to a collection of data points. How can one combine these two approaches? This book seeks to answer that question by presenting case studies that combine the macroscopic examination typical of economic history with the tightly focused analyses common in art history. I aim to show that these apparently incompatible approaches are, in fact, complementary. Data can better contextualize the stories of individual artists and objects; and paying close attention to the stories of these artists and their works can provide better insight into the individual choices and details that, in aggregate, become a general trend in the data.

In doing this, I hope to provide—as do Harrison and Cynthia White in their landmark book Canvases and Careers: Institutional Change in the French Painting World (1965)—a novel view of the nineteenth-century art world that sketches out its structural contours without losing sight of the individual painters and objects that existed within those structures. The nineteenth-century art world is particularly well suited for this type of approach for three reasons: the large scale of artistic production during the period, the extent to which this production is documented in preserved exhibition catalogs and other written sources, and the large quantity of nineteenth-century artworks that have since faded into obscurity or been lost—and therefore have largely disappeared from art historical narratives.

While these usable traces of past cultural production exist for a range of geographical settings, this book focuses on France, the United States, and England. Nineteenth-century art exhibitions in these three countries often (although certainly not exclusively) took place in large art academies or other centralized venues. As described at length in chapter 2, exhibitions were well documented in these three countries—and this documentation was fairly easily converted into usable data. The availability of these data combined with my own area expertise in nineteenth-century American and French art—as well as my familiarity with the history of British art—led to the geographic focus presented here. Ultimately, the broad scope of this book reflects the large scale of the data available, which in turn reflects the enormity of nineteenth-century artistic production.

Nonetheless, amid the wide-ranging potential of these large datasets, it is necessary to identify specific research questions that can be addressed with computational methods. Therefore, while chapter 2 introduces the datasets and fully embraces their breadth, the subsequent three chapters are case studies that show how a computational approach can add new evidence and perspectives to what I have identified as recurring important topics in the study of nineteenth-century art related to industrialization, gender, and the history of empire. In particular, chapter 3 focuses on the impacts of industrialization on art in nineteenth-century France; chapter 4 examines the profound effects of gender on American art and artists; and chapter 5 charts how art exhibited in England showed (or omitted) evidence of the country’s colonial expansion and enterprise.

Before delving into this new methodological approach and then the case studies, it is important to provide a couple of caveats. I am aware that the discipline of economics—and by extension economic history and the quantitative social sciences more generally—is not ideologically neutral. Economic theories and studies are often political tools. The discipline—beginning with the foundational work of Adam Smith—emerged at the same time as the contemporary capitalist system, and one of its core goals is to describe, understand, and predict how markets and people function in this system. For these and other reasons, art historians and other humanists may be resistant to an economic or quantitative approach to the history of art. Though this social science may, for some, evoke soulless data-driven analyses and capitalist agendas, I ask readers to be open to a reevaluation of modern economics and its value to the humanities. This book uses data, quantification, and economic theory as analytical tools that can address long-standing research questions in art history in totally new ways. Without these tools, it would be impossible to gain these insights. However, I want to make clear that just using quantification and data do not provide some sort of objective truth. Correlations presented do not translate to definitive causation. Rather, they simply provide a different perspective and new kinds of evidence.

This may, to some, seem like a profound lack of skepticism about the ideological underpinnings of the methodologies that I deploy. Instead, I view it as pragmatism in the name of novel inquiry. Economists and social scientists have been analyzing large datasets for a century; as digitization and new computing technology make it possible to engage with art historical data, it is counterproductive to ignore this disciplinary expertise for ideological reasons. I hope to present convincing examples of the ways in which economic and data-driven approaches are multifaceted and helpful. I will show (rather than tell) those who may be skeptical of these methods that they are, in fact, valuable complements to the study of art history. Furthermore, they do not supplant the qualitative humanistic core of the discipline.

This book bridges art history and economic history in two principal ways. First, it applies quantitative methods and statistical analyses typical of the social sciences to art historical subject matter—similar to what Prown did. To do this, it relies primarily on three new datasets about hundreds of thousands of artworks shown in nineteenth-century France, England, and the United States. These datasets are crucial, as it would be impossible to forge a quantitative history of art without this kind of information. Furthermore, these kinds of large datasets can help art historians address a phenomenon that potentially compromises our research: sample bias. Second, it draws on economic theory that seeks to explain how and why people act in a certain way, both in market settings and in broader society. Art historians are expert at chronicling and analyzing detailed information about the lives and careers of the artists, collectors, dealers, and institutions we study. However, we rarely step back and generalize about commonalities in the actions of all these subjects and how they reflect broad patterns of behavior as elucidated by social scientific disciplines like economics.⁷ Taking a cue from economics’ generalizing impulse and drawing on its theoretical conclusions can yield valuable art historical insight.

This introduction begins by laying out several key concepts in economics that provide touchstones for creating a data-driven history of art. Following these descriptions, I provide a brief literature review of earlier interdisciplinary efforts that have combined quantitative analysis and art history. The scholarship reviewed is from both the social sciences and the humanities. The final section of this introduction describes what makes certain kinds of art historical research questions particularly suited to this kind of inquiry. In some ways, what this book presents can be described as digital humanities, because it relies on computing technology both to generate and to analyze datasets. However, I believe the analyses presented here go beyond the digital because they take on humanities subject matter with a social scientific approach. To clearly identify my methods as a distinct blend of elements from economics, art history, and the digital humanities, I have opted to refer to this work as a data-driven history of art.

Sample Bias: Understanding the Limits of Art Historical Knowledge

Sample bias, a term frequently used by social scientists, emerges when the group of people, objects, or other entities that a scholar analyzes is both limited and misrepresentative of the entire population that the scholar is interested in studying. A classic example comes from political polling. Prior to the 1936 US presidential election, the periodical Literary Digest conducted a poll. It sent straw vote ballots to more than ten million Americans, which represented almost 8 percent of the population. Participants were mostly selected using automobile registration and telephone listing information. Over two million ballots were returned (about 20 percent of the number distributed), and the magazine published the results without any adjustments to the data generated by the responses. The poll predicted that the Republican candidate, Alf Landon, would win in a landslide. The opposite happened. Democratic incumbent Franklin D. Roosevelt won with an overwhelming majority. How could Literary Digest reach such incorrect conclusions? The answer is, in part, sample bias. By selecting the recipients of the straw poll from lists of automobile and telephone owners, the poll captured a greater number of wealthy Republican-leaning voters than of the generally poorer Democratic supporters. There was also systematic bias in who was more likely to return completed ballots. As follow-up surveys by the polling firm Gallup showed, more Landon supporters returned their ballots than Roosevelt supporters. The editorial team at Literary Digest drew incorrect conclusions because the sample of information it was working with was limited and misrepresentative of the population of interest—it was biased.⁸

How does this concept relate to art? In art history, sample bias is subtler than the example of the Literary Digest poll. The sample of artworks that can be the subject of art historical study is necessarily limited. Artworks are damaged and lost; some pieces are not saleable and therefore never end up in a private collection, and certainly not a public one; museums have limited space and must be discriminating in what they choose to collect. The end result of this process is that only a limited number of artworks ever created will be preserved and available for study. A further narrowing of the sample occurs when disproportionate attention is given to particular artists whose work is well known and accessible—notably, those artists who are currently famous, have ended up in major museum collections, or are featured in typical art history survey courses.

Art historians therefore necessarily study a sample of the entire population of artworks. However, it is difficult to know the specific ways in which that sample may diverge from or reflect the historical population of artworks. First, the qualitative methods and close looking that are the fundamental components of art history demand that scholars limit their focus to a narrow sample of artworks. These approaches are designed for focused analyses, not population overviews. Second, there is limited information available about artworks that no longer exist, are unlocated, or have faded into obscurity. When the focus of a method of inquiry is, quite correctly, looking at the artwork, it is hard to account for artworks that one cannot see or study.

One contribution of this book is to provide new data about these forgotten artworks. It repurposes records of past cultural production—notably, historic exhibition catalogs—and reintegrates them into art historical consideration. The transcription, digitization, and statistical analysis of lists of artworks produced and displayed at institutions during the nineteenth and early twentieth centuries provide vast and previously untapped historical sources. As chapter 2 demonstrates, there were hundreds of thousands of works made and shown during the long nineteenth century in a handful of venues in France, the United States, and England alone. While these sources provide only trace records—mostly lists of artists, titles, dates, and media—these data nonetheless provide a new opportunity to compare careful histories focused on select artists and works of art to population-leve...

Table of contents

• Cover Page
• Title Page
• Copyright Page
• Contents
• 1. What Is a Data-Driven History of Art?
• 2. The Historical Data of the Art World
• 3. Between City and Country: Industrialization and Images of Nature at the Paris Salon
• 4. Why Have There Been No Great Women Artists? Artistic Labor and Time-Constraint in Nineteenth-Century America
• 5. Implied But Not Shown: Empire at the Royal Academy
• 6. Conclusion
• Appendices
• Acknowledgments
• Notes
• Bibliography
• Index