It may still be the case that historians, as a whole, are averse to using databases of digitized primary sources in their research. My impression is that this is rapidly changing, however. This impression is admittedly unscientific and based only on the scholarship that I read. My perceptions may also be skewed by the fact that I myself have found digital databases useful in my research, as illustrated by my last post on a Lincoln quote and my previous series on John Brown’s Timbuctoo.
Still, in at least one field—the history of the early American republic—there is lots of evidence that scholars already see digital databases as crucial to their research. Recent historians of the early republic even seem eager to deploy keyword searches and share their digital findings. In this post, I’ll illustrate what I mean by citing some recent examples of how historians in my field are using proprietary digital databases.
For the past year or so I’ve been keeping an incomplete but running list of articles in the Journal of the Early Republic (the official journal of SHEAR) that make explicit use of proprietary databases published by companies like ProQuest, NewsBank and Accessible Archives. By sharing these examples, I hope to provide a quick snapshot of some of the actual practices of historians who use digital databases, particularly historians who don’t seem to identify primarily with the field of digital history or digital humanities. Finally, at the end of the post, I’ll explain why I think historians in my field could benefit from a central online repository that makes information about these databases accessible and keeps track of differences among them.
In many ways, historians of the early nineteenth century are in the best position to exploit digitized primary sources. Copyright laws being what they are, most of the full-view books available on Google Books fall in our period. Database companies have also been extremely successful at digitizing nineteenth-century newspapers and ephemera. Important historical collections that once existed on microfilm now exist in digital form thanks to proprietary databases like ProQuest’s American Periodical Series Online and Readex’s Archive of Americana. Given the availability of such collections at many academic libraries, it’s not surprising that historians of the early republic have begun using these databases and citing them in their work.
Within the pages of the Journal of the Early Republic, the use of such databases has so far run the gamut from casual mentions of keyword search results to more ambitious efforts to build arguments around such results.
Examples of the more casual references include John L. Brooke’s 2008 presidential address to SHEAR, “Cultures of Nationalism, Movements of Reform.” After a paragraph arguing that “prayers by a local minister seemed almost universal” at celebrations of the Fourth of July and Washington’s Birthday, Brooke included a footnote stating that “my comments on religion and national celebration are based on the results of searches in Early American Newspapers and Gale 19th Century United States Newpapers databases.” The year before Brooke’s address was published, Caroline Winterer’s introduction to the Journal’s Spring 2008 roundtable on Mary Kelley’s book Learning to Stand and Speak also included a passing reference to results of a keyword search, but with more specific information about the search performed and the results. After noting the previous scholarly neglect of Kelley’s topic–female academies in the early republic–Winterer wrote that “numbers alone can show that this neglect of the female academies is entirely undeserved. Search the term female academy in the hundreds of American magazines that make up the American Periodicals Series online database and you retrieve exactly 1,131 hits for the period 1790–1860.” Similarly, an endnote in Daniel A. Cohen’s Spring 2010 article “Making Hero Strong” noted that “a keyword search of the phrase ‘story paper’ in ProQuest’s American Periodical Series Online 1740–1900 suggests that the term had entered into common usage by the late 1850s, if not earlier.”
In each of these cases, Brooke, Winterer, and Cohen used “hits” to back up points that were secondary to their arguments. But recent issues of the Journal of the Early Republic have also included several articles that make more extensive use of proprietary databases. Here’s a run-down of examples:
- In 2008, Carol Lasser analyzed language in antislavery newspapers to show that around 1840 there was a steep decline in the “sexualized imagery” of slavery that abolitionists had used frequently between 1834 and 1839. Using keyword searches in a subset of newspapers available in the American Periodical Series Online, Lasser identified the number of instances in which the words “rape, ravish, amalgamation, adultery, and the truncated licentious” appeared in the same paragraph as slave (and in the same article as either United States or America). She then displayed her results on a graph, with numbers of hits on a y-axis and years on an x-axis. After using this and two related graphs to demonstrate the “dramatic rise and subsequent decline” of such language, Lasser connected this pattern to contemporaneous changes in abolitionist strategy and the roles that women played within the abolitionist movement.
- In 2009, Mark Schmeller included a similar graph at the beginning of an article on the partisan origins and uses of the term “public opinion” in the 1790s. To demonstrate that in the 1790s a “chorus of newspaper writers … had just begun to fortify their arguments with references to public opinion,” Schmeller performed a keyword search of “public opinion” in the database Early American Newspapers, Series I, published by Readex. He then graphed the results with occurrences per 100 publications on the y-axis and years on the x-axis. For Schmeller’s argument, this timing of the uptick in occurrences was important to his larger argument that discourse about the meaning of “public opinion” was shaped by debates over political economy and public finance occurring at the same time.
- In 2009, Amanda Bowie Moniz performed searches in a few titles in Readex’s America’s Historical Newspapers for instances of the word drowned in order to find and catalog instances of drowning in the 1770s and 1780s. These results became one part of her argument that the rise of increasingly cosmopolitan humane societies for the resuscitation of drowning victims were not driven primarily or only by an increase in drowning incidents.
- Also in 2009, Nathan Kozuskanich made a forceful argument for the significance of digital research methods by suggesting that they can help settle legal disputes about the “original intent” of the Second Amendment. Kozuskanich searched for the exact phrase “bear arms” in a range of years in Readex’s digital version of the famous Evans bibliography of early American documents, in Readex’s Early American Historical Newspapers, and in the Library of Congress’s U.S. Congressional Debates database. In the combined results from all of these searches, which turned up almost 400 relevant hits, only about a dozen documents “do not use an explicitly military context when discussing bearing arms,” undermining the individual-rights interpretation of the Second Amendment’s original intent. Kozuskanich then went on to argue that close reading of his results also showed weaknesses in both of the understandings of the amendment’s original intent that prevail today.
- Again in 2009–a banner year for keyword searching in the Journal–Matthew Rainbow Hale used searches in Readex’s America’s Historical Newspapers to argue that “even after the increase in newspapers is taken into account, [temporal and journalistic] key words and phrases–including ‘millennium,’ “accelerated,’ and ‘rumor’–appeared more frequently between 1793 and 1795 than in two other eras (1774–77 and 1787–89).” Hale provided a table that listed the searches he performed on separate rows. Each column, headed with years, gave figures “representing the number of references” to the word or phrase “divided by the number of newspapers in the Readex database published during the selected years,” rounded to “the nearest thousandth.” He then used this evidence as part of a larger argument that the radicalization of the French Revolution in these years “generated a new level of anxiety [in the United States] regarding political time and news.”
Now, one could probably enumerate other recent examples of historians of the early republic using these and other digital databases. And it may be foolhardy to generalize about a set of articles as diverse in method and argument as these.
But two things strike me as I look over these cases:
First, it’s clear that many historians are not only using proprietary databases, but also wish to embrace the opportunities that the scale of each database make possible. Given the number of heated discussions in some other fields about whether text mining is a legitimate way of “reading” texts, it’s noteworthy that in each of the cases of above, the data is presented without much wringing of the hands. To be sure, many of the articles listed here are conscientious about mentioning the limits of their methods and heading off objections to their samples and searches. While Winterer described the 1,131 hits she found for the term female academy as an “extraordinary harvest,” she also admitted that this was a “crude measuring device.” Lasser provided additional analyses of her data to offset possible objections that changes in the number of abolitionist publications did not account for the patterns she was finding. Moniz explained her choice of the search term drowned by explaining that her own comprehensive survey of a New York newspaper not included in her digital search results revealed that most reports of drowning incidents used that phrase. Kozuskanich admitted that his numbers did not take into account the appearance of his search term multiple times in the same document. Each article also combined arguments made from keyword searches with more traditional close readings of documentary sources.
Even these qualifications and concessions, however, can be seen as indicators of a general optimism among the authors about the utility and defensibility of looking for quantitative patterns in digital databases. Historians are eagerly venturing into this new digital age without feeling like it means they have to leave old methods behind.
Yet It’s equally clear that historians have not yet developed clear-cut conventions for describing their searches, citing the databases used, and reporting their results. In these cases, URLs meant to point to the homepages of the proprietary databases sometimes included the “ezproxy” suffixes included in these URLs by their home universities. Perhaps more significantly, the articles employ multiple conventions for formatting search terms. Occasionally the search terms were italicized, while in one instance (Schmeller’s “public opinion”) they were placed in quotes. My initial impression as I compiled this list was that single word search terms were always italicized, while phrases were put in quotes, but some two-word phrases were italicized without quotes.
Except when the author specified that the search was for “the exact phrase,” these formatting conventions could sometimes leave unclear, at first glance, whether searches for two words were Boolean searches (United AND States) or exact phrase searches made with database-specific delimiters (“United States”). Given that the default settings for searches in different databases treat Boolean and delimited phrases differently, knowing exactly how terms were inputted will probably be important to many readers of these articles. There was also some variation in the way truncated terms were represented; in Hale’s table, for example, one of the search terms listed is false report(s), but it is unclear whether the results on this line include the combined results of two searches (for false report and then false reports) or a search for false report which thereby included, according to database-specific conventions, all instances of the pluralized phrase as well. There was also no uniform convention for how to cite the proprietary database or its publisher (which was in two cases left out). And last but not least, while Hale provided the month and year in which he performed his searches, other authors did not.
To illustrate the stakes involved in these seemingly slight differences, consider Winterer’s use of a search for the term female academy. Today, when I performed a search for female academy in the American Periodical Series Online, limiting the search to the period between 1/1/1790 and 12/31/1860, I retrieved 1,435 hits–around 300 more than Winterer reported. I also got a different number of hits if I searched only in “document text” and not in “citation and document text.” In this example, I tried my search for female academy both with and without quotations around the phrase and got the same number of hits. But when I performed the two searches (female academy and “female academy”) on America’s Historical Newspapers, published by Readex, I got over 100,000 results for one and less than 10,000 for the other.
Winterer’s point is not significantly changed by a difference of 300 hits, but other arguments might be impacted by such variations. And these variations are important primarily because they present difficulties to a reader wishing to evaluate articles that use keyword searches.
If these articles do indicate a trend towards such methods among historians, that means that all historians–whether they use such methods themselves or not–will increasingly be placed in a position of needing to review and evaluate such methods. I see this as a good thing, but one problem I foresee for historians is the difficulty of keeping track of differences between databases. To evaluate search results–especially when an argument rides on differences between relatively small numbers of hits–it will become increasingly important to know specific features of databases that may not always be reported by authors: do default searches include “fuzzy” hits? is the text being searched created by Optical Character Recognition or human transcription? how many sources does the database include, and over what chronological and geographical range? how often is the corpus changed or updated, if at all?
To address some of these issues, Lasser’s endnotes helpfully provided a link to general information about the American Periodical Series provided by ProQuest. But this convention, too, was not uniform across the examples I’ve surveyed here. In these articles, and probably in future ones, it will fall to readers to seek out relevant information about the databases themselves. But such a search is not easy or intuitive, given that different database companies present the information in different ways, on different pages, and with differing degrees of transparency. My rough sense from perusing pages like the one linked above is that companies’ descriptions of their products are usually aimed more at buyers than at users. Likewise, comparisons of databases that can currently be found on the open web are written by librarians and purchasers interested in getting their money’s worth. Where does that leave the historian who instead wants to evaluate search results being used as evidence for historical arguments?
This is one of the questions I’m hoping to be able to talk about at THATCamp Texas next week. More specifically, I’m wondering whether it might be useful and feasible to create some resource that compiles relevant information about the proprietary databases historians use most so that readers (and also writers) can quickly get a sense of the lay of the land at a particular database. What I’m imagining is something like a SHERPA/Romeo site, but geared towards the description of search functions and interfaces instead of copyright policies. The aim of such a site would not be to shut down digital methods like the ones employed in these articles or serve as a gatekeeper for publication. On the contrary, I think such a site would make historians more comfortable with such methods and would help build on the existing momentum towards their use. Such a site would be useful to authors and researchers as well as readers, I suspect. But does such a site exist? What information would make such a site be useful? What costs and problems would be involved in building one? How might it be pitched so as to encourage, rather than discourage keyword searching?
John L. Brooke, “Cultures of Nationalism, Movements of Reform, and the Composite–Federal Polity: From Revolutionary Settlement to Antebellum Crisis,” Journal of the Early Republic 29.1 (2009): 1-33 § Caroline Winterer. “Women and Civil Society: Introduction,” Journal of the Early Republic 28.1 (2008): 23-28 § Daniel A. Cohen, “Making Hero Strong: Teenage Ambition, Story-Paper Fiction, and the Generational Recasting of American Women’s Authorship,” Journal of the Early Republic 30.1 (2010): 85-136 § Carol Lasser, “Voyeuristic Abolitionism: Sex, Gender, and the Transformation of Antislavery Rhetoric,” Journal of the Early Republic 28.1 (2008): 83-114 § Mark Schmeller, “The Political Economy of Opinion: Public Credit and Concepts of Public Opinion in the Age of Federalism,” Journal of the Early Republic 29.1 (2009): 35-61 § Amanda Bowie Moniz, “Saving the Lives of Strangers: Humane Societies and the Cosmopolitan Provision of Charitable Aid,” Journal of the Early Republic 29.4 (2009): 607-640 § Nathan Kozuskanich, “Originalism in a Digital Age: An Inquiry into the Right to Bear Arms,” Journal of the Early Republic 29.4 (2009): 585-606 § Matthew Rainbow Hale, “On Their Tiptoes: Political Time and Newspapers during the Advent of the Radicalized French Revolution, circa 1792–1793,” Journal of the Early Republic 29.2 (2009): 191-218.
Image Credit: Screen shot of an article from Genius of Universal Emancipation, April 10, 1823, American Periodical Series Online