The following remarks (here slightly condensed and reformatted) were prepared for a talk about the role of digital humanities scholarship in the classroom, although you’ve got to get to the very end before you see some of the pedagogical scenarios I propose. I’m posting these remarks here, also, because I haven’t yet blogged about some of the social network analysis and visualization techniques I’ve been exploring of late, and I’m eager to hear others’ comments and questions– and suggestions– about the project.
At the outset, I should also thank David Sewell, with the University of Virginia Press’s Rotunda imprint, for granting me access to the XML files of The Papers of Thomas Jefferson Digital Edition. The textual analysis that I describe in this talk would have been virtually impossible without the support of the UVA Press.
How We Know What We Know: Thomas Jefferson, James Hemings, and the Task of the Scholar (and Teacher) in the Digital Age
In preparing these remarks, I found myself returning to a line from the introduction to SpecLab, Johanna Drucker’s 2009 meditation on the digital humanities center that she helped to create at the University of Virginia nearly ten years earlier. In attempting to define the digital humanities for her readers, she explains that the field does not simply entail the creation of “new electronic environments for access to traditional or born-digital materials. It is the study of ways of thinking differently about how we know what we know and how the interpretive task of the humanist is redefined in these changed conditions” (xii).
What continues to resonate with me about Drucker’s definition is how she conceives of the digital humanities both in terms of its epistemological function—about “how we know what we know”—and in terms of its function for us as critics—and I will add here, as teachers as well. It’s my hope that with the example I’ll provide today, I will demonstrate how the question of “how we know what we know” has guided me from my dissertation research to my own digital humanities scholarship, and how, over the course of that investigation, it has prompted me to inquire about my role as a literary critic, working in the archive of American culture today. I also hope that with this example, I will illustrate how I encourage my students—as well as the faculty members I work with in my capacity as an Instructional Technology Fellow—to question the function of their scholarship, and to examine their roles as humanists in what Drucker would call the “changed conditions” of the digital age.
My dissertation demonstrates how a shared cultural language of food– or, more accurately, eating– emerged out of the transition from colonial rule to the early republic, and how that language transformed, over the generations that followed, into a national consensus about the interdependence of the cultivation of the American palate and the cultivation of virtuous citizenship. I argue for a conception of a distinctly American sense of taste, one that is both gustatory and aesthetic. I show how this composite sense of taste expresses the ideals associated with the nation’s founding, and at the same time, incorporates its enduring contradictions of race, gender, and class. One of my dissertation chapters is about Thomas Jefferson, who was one of America’s first great gastronomes, as well as one of its founding fathers. And of course, as is widely known, he was also a slaveholder.
In fact, when Jefferson was appointed Ambassador to France, in 1785, he took an enslaved man, James Hemings, with him to Paris, and enrolled him in culinary school. James Hemings was the older brother of Sally Hemings, recognized now as the woman with whom, whether consensually or not, Jefferson maintained a lifelong relationship, and who bore six of his children. Because of the outstanding archival work of Annette Gordon-Reed, whose biography of the Hemings family was published in 2007, we also know that, for instance, while in Paris, Hemings employed a personal tutor in order to learn to speak fluent French. And another thing Hemings learned in Paris was what it might mean to be free.
During Jefferson’s tenure as Ambassador, France adhered to what was called the “Freedom Principle,” which held that any enslaved person who set foot on French soil would be, from then on, considered free. Hemings could have invoked this principle of freedom, and liberated himself from bondage, but he chose not to. Instead, he continued to hone his gastronomical skill, and eventually became the head chef in Jefferson’s Parisian residence. Gordon-Reed suggests that Hemings most likely made a verbal agreement with Jefferson to return to America, whereupon he would receive his freedom. The only written evidence for this, however, is a document from four years after their return, a formal, although not legally-binding agreement that establishes the conditions for Hemings’s eventual emancipation.
The agreement notes the “great expense” of “having James Hemings taught the art of cookery.” It expresses Jefferson’s desire to “befriend him, and to require from him as little in return as possible,” before stipulating that Hemings must instruct a replacement cook before he can be freed. In essence, the agreement proposes that Hemings ensure the legacy of his culinary knowledge in exchange for his personal liberty. Jefferson’s conditional proposition—to which Hemings had no choice but to consent—exemplifies the form of barbarism that Saidiya Hartman identifies, in Scenes of Subjection, as evident not only in the “constitution of slave as object but also in the forms of subjectivity and circumscribed humanity imputed to the enslaved” (6). Indeed, Jefferson’s measured tone and offer of friendship illustrate, in stark relief, the incontrovertible authority of Jefferson as master, and the resultant subjection of Hemings as slave. Jefferson characterizes himself as a benevolent force of liberty, but his concern with the practical implications of Hemings’s release reveals the ways in which his heightened valuation of the “art of cookery” takes precedence over the foundational rights of the republic. In regard to the argument I make in my dissertation, about the contradictions enforced through the cultivation of American taste, this document exemplifies—to the highest degree—how the daily, lived experience of slavery enters into—and ultimately disrupts—the more polished narrative that Jefferson seeks to promote about the values and ideals inherent in American taste.
But what I want to focus on, today, is not the shadow cast by slavery on Jefferson’s conception of American taste. Rather, I want to discuss how the shadow cast by James Hemings, as I first encountered it in the Jefferson Digital Archive, prompted me to explore “how I knew what I knew” about Hemings’s life, and how I might employ a set of digital tools and methods—in particular, natural language processing software and social network visualization techniques—in order to know more. I’ll then address the second aspect of Drucker’s definition of digital scholarship, about how the “interpretive task of the humanist” is redefined in the digital age, by way of my students’ response to my project, and I’ll suggest how a related set of digital tools might be employed in the classroom in order to further expand the bounds of humanistic inquiry.
When you search for “James Hemings” in the Jefferson Digital Archive, you get no results. In the screencap on the left, you see a vivid instantiation of what Michel-Rolph Trouillot, the eminent historian, has called the “silence” of the archive. More recently, though, scholars from across the humanities, such as the literary critics Stephen Best and Sharon Marcus, the sociologist Avery Gordon, and the archivist Jeannette Bastian, have each shifted their focus from identifying and recovering these silences, to a new focus, instead, on animating the mysteries of the past. Best and Marcus, for instance, tell us we must learn to see shadows in the archive, shadows such as Hemings, as “presences, not absences, and let ghosts be ghosts, instead of saying what they are ghosts of” (13). Their point, which is an important one, is that we must find a way to acknowledge the silences of history without reinforcing the damaging notion that African American voices from before emancipation—not just in the archival record, but the voices themselves—are silent, and irretrievably lost.
Best and Marcus propose a theoretical solution to archival silence—one that I find quite compelling—but I would like to propose a practical solution that might complement their aims: let’s visualize these absences, let’s animate these ghosts of the past. If few documents exist that were written by or to James Hemings, what about those—like the emancipation agreement—that were written about him.
Using Protovis, a javascript-based toolkit from Stanford, that allows users to present textual data in various visual forms, I was able to create the visualization to the left (click to enlarge). What I have created is called an arc diagram. It visualizes the people with whom Jefferson corresponded about James Hemings and his family. This is what sociologists call a social network–a concept that predates our current understanding of social networking websites, like Facebook, but in fact works in a similar way. I compiled my social network data by performing a series of searches in The Papers of Thomas Jefferson Digital Edition, a newer, more complete site, launched in 2009 by the University of Virginia Press’s Rotunda imprint, based on the thirty-six volume (and counting) Princeton University Press print edition. Because the Digital Edition allows users to search the Papers’ editorial notes, as well as its content, I was able to identify some—but not all—of the correspondence that mentioned members of the Hemings family.
I discovered that Hemings’s last name was spelled at times with two Ms and at other times with one. What was more troubling, and also technically problematic, was that because of Hemings’s status as a slave, he was more often referred to by his first name only. Not only are there many Jameses mentioned in the Jefferson archive, but James Hemings was also called Jamie, Jimmy, and even Gimmé while in France. Although in my search, I was able to discover a range of documents that referred to Hemings, I was reliant on the editors of the Papers to have reviewed each letter, one by one, in order to identify any references to Hemings, and then for them to have recorded the reference– with first and last name– in the accompanying editorial note. As one might imagine, this process is an intensive one, and the papers’ editors have been working on the project since 1950. One volume of letters is published in print and online every few years. My last inquiry revealed that the digital Papers currently includes material through 1801. In that year, Jefferson was fifty-eight. Jefferson lived to be eighty-three. And his papers’ editors still have approximately twenty-five additional volumes of his correspondence to annotate.
Which is all to say that while this visualization may be incomplete, it is still quite significant. You’ll notice that I’ve arranged the correspondents in groups—indicated by the different colors—from left to right, Jefferson and his family, Jefferson’s political correspondents, his hometown friends, his correspondents in France and abroad, his enslaved staff, his plantation overseers and free plantation staff, and on the end, people about whom I could find very little (or in some cases, no) biographical information. An arc connecting two names indicates that they corresponded. I have not distinguished between the author and recipient of each letter, since this is a preliminary rendering, but the width of the arc indicates the frequency with which they corresponded. So because these are the Jefferson Papers, all arcs connect to him. You’ll notice that darker, wider arcs connect to Nicholas Lewis, Jefferson’s neighbor in Virginia; George Jefferson– Thomas Jefferson’s Virginia agent, although apparently not a close family relation; and then to Richard Richardson, who worked as a plantation overseer at Monticello. Presumably, Jefferson corresponded with each of these men about the services that Hemings provided on the plantation. And this visualization points to the record of the labor—the difficult, daily labor—that Hemings was required to perform.
The fourth wide arc, though, the one that connects to William Evans, is—socially speaking—an outlier. Evans ran an inn outside Baltimore, and he wasn’t even a close friend to Jefferson. When you search for Evans in the Digital Edition, you turn up a single, sparse chain of correspondence from 1801, which was eight years after Jefferson recorded his agreement with James Hemings, and five years after Heming’s legal emancipation. In the first of these letters, Jefferson writes to Evans: “You mentioned to me in conversation here that you sometimes saw my former servant James.” Jefferson asks that Evans “send for him & tell him I shall be glad to receive him,” before shifting the focus of his inquiry to another man, Francis Sayes. According to the letter, Sayes also “lived with [Jefferson] formerly,” but little else about Sayes– including whether he was another slave, or a free man– is currently known. Sayes “had begun to drink a little before he left me, & I fear he continues it,” Jefferson reports to Evans. But more problematically, at least as far as the eternally financially-preoccupied President was concerned, Sayes had expressed repeated interest in seeking work at Jefferson’s new Washington residence. “He was an affectionate & honest servant to me, which makes me unwilling to reject him absolutely; and yet the fear of his drinking and of his getting his family into distress by removing them, induces me to wish rather that he would decline the thought.” Jefferson explains this to Evans, and then he makes a direct request: “Should he be with you, or fall into your way, I would thank you to discourage him.”
The multiple forms of communication mentioned to in this letter, alone, underscore just how small a section of the past is preserved in even the largest, and most exhaustive archive. At the start of the letter, Jefferson makes reference to a prior “conversation.” He then asks his innkeeper friend to “tell” Hemings a message, which emphasizes the complex trail of communication between men with permanent residences and those who were more itinerant, as well as between men and women who regularly sent letters, and those who, because of any number of cultural and political strictures, were unable or unequipped to write. In the letter to Evans, Jefferson also refers to the fact that his other former employee, Sayes, “says that his wife has good custom in Baltimore as a washer,” information that was either communicated orally, during a prior meeting, or quite possibly in another letter that was not simply preserved [emphasis added]. And in contrast to the specific message that Jefferson begins the letter by asking Evans to convey, at the end of the letter, Jefferson asks Evans to use his own words to “discourage” Sayes from seeking future employment at the President’s house.
As a result of this convoluted exchange, neither Hemings nor Sayes ever arrived in Washington. Evans was successful in convincing Sayes not to contact Jefferson, and according to Gordon-Reed, Hemings most likely took offense that Jefferson had not attempted to communicate with him directly, and never spoke to Jefferson again. In fact, Hemings’s response to Jefferson, conveyed in another letter from Evans, would be his last recorded sentiment. An eight month gap in the correspondence between Jefferson and Evans ensues. The subsequent—and final—exchange between Jefferson and Evans, from November 1801, confirms the “melancholy circumstance” of Hemings’s suicide.
—
We need not make the ghost of James Hemings stand for something, as Best and Marcus caution. To be quite certain, the ghost of James Hemings means enough. But what we can do is examine the contours that his shadow casts on the Jefferson archive, and ask ourselves what is illuminated, and what remains concealed. In the case of the life—and death—of James Hemings, even as we consider the information disclosed to us through Jefferson’s correspondence, and the conversations they record—we realize just how little about the life of James Hemings we will ever truly know. Was there some way I could animate this mystery, I then asked myself?
In view of not only the multiple modes of communication, but also the multiple voices that are recorded in the Jefferson archive, I decided to try to visualize the men and women who were mentioned in the content of the letters (left; click to enlarge). Technically, this was a much more challenging task. I contacted David Sewell, at The University of Virginia Press’s Rotunda imprint, who sent me the digitized versions of the Jefferson Papers in XML form. I then extracted the text of the letters from those files, and ran it through what’s called a named entity recognizer—software that derives from the field of computational linguistics that is able to identify, or “recognize,” sequences of words in a larger text that represent so-called “named entities,” mostly people and places, and some things. The names that you see on the bottom of this visualization are people who are mentioned by name in the letters, as determined by the named entity recognizer. I went through the results afterwards by hand, in order to eliminate the errors and duplicates that I could recognize—such as the variations in James Hemings’s name mentioned earlier, or changes in name that occur during a person’s life, such as in the case of Martha Jefferson, Jefferson’s oldest daughter, who was referred to as Patsy as a child, and became Martha Randolph after marriage. In preparing this visualization for today, I also eliminated about a hundred names of people who were mentioned in the set of letters only once, so that the diagram could fit on one screen. I then wrote my own co-appearance analysis script, in Python, in order to determine which people were mentioned in the same letter as each other, and how many times those people appeared together. And then I formatted the data to be displayed using Protovis, as before. So the arcs shown here are generated from the same set of letters as in the previous diagram, but the relations among the people mentioned are much more complex.
Significantly, the arcs that link Jefferson to his slaves are much wider than those that link him to many of his correspondents, suggesting the degree to which he relied on his enslaved staff to implement his various directives—buying things, selling things, even telling them things—in fact, it would appear, to an extent greater than the people with whom he directly corresponded. The visualization I’ve created brings these dependencies to the surface, but it still does not provide answers to the questions of what these people said in their conversations, where they went in order to conduct their transactions, or how they truly experienced their everyday lives.
When I showed this visualization to some of the students I work with at CUNY, their suggestions for how I might improve the project proved this point. “What if you made it draw the arcs over time,” one student offered. “Then you could see which people were more important to Jefferson, and when,” she said. “What about if you could do this for all of the letters in the archive,” another one suggested. And in fact, that’s a direction I’ve also considered, and one that I’d like to begin to implement. “Could you put it on a map?” another asked. “These letters were sent to places, right?” Or, my favorite: “Why don’t you make those lines link back to the letters,” this one student said. “Then you could find out what the letters said.”
All of the suggestions my students offered—framed pragmatically, and offered as friendly advice—confirmed for me the ways in which digital tools and methods—and not just the advanced techniques I’ve employed here, but also blogs, wikis, podcasts, maps, what have you—bring the process of humanistic inquiry to the surface. My students quickly recognized the information presented by my visualization, but they almost immediately began to ask how they could learn more.
This is the process I aim to bring to the classroom, with the way in which I incorporate digital methods and tools. These tools allow us, as Johanna Drucker explains, not only to create digital versions of prior research, or digital frames for presenting scholarly analysis; they also open up to us—and to our students—the possibilities of humanistic inquiry. I’d be happy to discuss some of the projects I’ve worked on with students—and faculty—in the past, that have ranged from timeline projects to travel blogs, video documentaries to wikis. But I want to end, briefly, by showing how some easy-to-use, web-based text-analysis tools, might allow students to participate in—or even to initiate—work similar to that which I’ve shown today.
Here’s the text of all the Hemings letters cut-and-pasted into Wordle (left). Wordle lets you make word clouds like the one to the left, with the size of the word corresponding to the number of times it appears. (It selected the “olde” font automatically, based on the archaic language that my pasted text contained). But you can see even here, larger words like “time,” “money,” and “work.” I’d love to show this to students who’d read a bit about James Hemings, and ask them what they thought it implied.
Here’s another example, using the Google Ngram Viewer (left). This is a comparison of the relative frequency of occurrence of the names “Jefferson” and “Hemings” in the period of time from Thomas Jefferson’s birth to the present. It’s mostly as you would expect, for a comparison between a US President and an man he enslaved, but what about those small bumps, clustered around a few distinct time periods? The most recent one, I would guess, was in response to the DNA testing that confirmed that Jefferson had, in fact, fathered children by Sally Hemings. But wouldn’t it be wonderful to send students on a historical quest, in order to determine what the earlier bumps in frequency corresponded to.
Or, finally, here’s another cut-and-paste tool, IBM’s Many Eyes, that allows you to produce what’s called a word tree. A word tree is a map of what words and phases follow from a particular word that you’ve selected. So here I started by searching for James, and just by reading down the resultant connecting phrases, you can see how James Hemings was referred to sometimes as “James,” and sometimes as “James Hemings.” Then you can also see in what contexts he was mentioned—“shall go with me to Monticello,” “is now of the age of thirty.” From this, you get as clear a sense as any of the life James Hemings lived.
In an essay, “In Praise of Pattern,” Stephen Ramsay, one of the pioneers in visualizations of literary texts, whose own digital scholarship centers on the statistical analysis of scene structure in Shakespeare’s plays, underscores the importance of aligning digital tools and methods with what he calls the “imperatives of humanistic inquiry.” What I hope I’ve demonstrated, today, is an example of how digital tools, when aligned with interpretive questions, can open up our eyes as scholars, and as teachers, to the possibilities of humanistic inquiry. I hope I’ve shown how digital tools challenge us to articulate—in the full range of media, and the cultural forms that we encounter in our everyday lives–how we know what we know, and also what else we might learn. So I’ll end there, and with that, I invite your questions, and your ideas.
