Moving Beyond Anecdotal History

Moving Beyond Anecdotal History
Gibbs, Fred, fwgibbs@gmail.com George Mason University,
Now almost fifty years old, Walter Houghton’s seminal work, The Victorian Frame of Mind, 1830-1870, has influenced generations of scholars of the nineteenth century and remains the primary introduction to Victorian thought that every student in the field reads. From a close reading of famous Victorian writers such as John Stuart Mill and Thomas Carlyle, Houghton argued that the Victorians were characterized by specific, common personality traits such as optimism, hero worship, and earnestness. Houghton believed that these traits were visible in the rise (or decline) in the use of particular words and phrases, such as an increasing use of “light,” “sunlight,” and “hope” as illustrative of their optimistic world view.

Despite the enormous impact of The Victorian Frame of Mind on generations of scholars across the humanities, it has not been accepted uncritically. Many concerns stem from Houghton’s myopic textual methodology: generalizing the character of a people—“the Victorians”—from the words of a select few. Although Houghton cites hundreds of primary sources in his bibliography, his book has been characterized as anecdotal, elite intellectual history. Despite such criticisms, Victorianists have been able neither to thoroughly assess the general validity of Houghton’s theses nor to offer alternatives.

New digital tools and the vast digital library of Google Books now allow us to conduct a comprehensive survey of Victorian writing—not just the well-known Mills and Carlyles, but tens of thousands of lesser-known or even forgotten authors—to test whether the Victorians truly did use the kinds of words and phrases that Houghton claimed they did. Did metaphors of hope actually increase in real terms between 1830 and 1870? Or was this only true for the dozen prominent writers he examined for his chapter on optimism? How can we complicate Houghton’s characterizations and understand change over time through the vast index of Google Books? Can we refine the timeline for the emergence of his characteristics, moving beyond the disturbingly neat, rounded-year boundaries he set for his book? How can we correlate historical events with disturbances in the linguistic data? To what extent can we separate cultural history from printing history with a large corpus of digitized literature?

Dan Cohen, my colleague at the Center for History and New Media, and I have begun work on a project to answer these very questions. With the help of a Google Digital Humanities Grant, we’re attempting what Franco Moretti calls a “distant reading” of the Victorians. My paper will explain how we’ve gone about querying the data available through Google Books, how we’ve been able to make sense of and interpret the results, and how we’ve dealt with the messiness of the data. I hope to solicit feedback about our methodologies and conclusions as part of a larger discussion about how the Google Books corpus (and similar datasets) could be made more usable for large-scale data mining projects relevant to the diverse research interests of the audience. How far can we push our methodologies beyond testing certain theses in order to allow the texts to speak for themselves?

In terms of the Victorians, some preliminary results have proven quite intriguing. For example, the number of books published with “universal” in the title declined steadily throughout the century, but earlier than most interpretations in the secondary literature point out. A look at published titles suggests that the terms “God,” “Christian,” and “Bible” follow rather different contours, though explanations are not immediately apparent. Similarly, how can we explain the striking publication parallels between the terms “belief” and “Aristotle”? The median number of titles that use the word “hope” does not significantly change between 1830 and 1870: does this cast some doubt on Houghton’s characterizations? Or does it simply indicate that book titles are not an accurate gauge of popular sentiment? We have not yet been able to examine the full texts from the publications that are being counted, but we hope to do that soon. To what extent will a more sophisticated linguistic analysis of the full texts reinforce or contradict what the titles alone tell us? Does this have implications for similar large-scale research methodologies?