Cite Unseen: Methodological Essay

This page offers a brief discussion of the dataset that I explore in my paper “Cite Unseen: Mapping Information Networks in the Revolutionary Atlantic.”

Building the dataset

For Canadian newspapers, gathering direct citations was a relatively straightforward process. It involved moving steadily through the pages of the Royal St. John’s Gazette, Halifax Gazette, Quebec Gazette, and Montreal Gazette. These are not the only papers published in Canada in this era, but they are among the only long-running papers that have been well-preserved and were accessible to me. They are reasonably good reflections of the two centers of printing in Canada: the former two papers were located in the so-called “maritime provinces,” while the latter pair were in Lower Canada (today’s Quebec).

For newspapers in the United States and the colonies that became the United States, my citation-gathering process was somewhat more complicated. I found citations in the America’s Historical Newspapers database using a two-step process. First, I executed a search for headings that used the signal phrase “from the.” Most direct citations used this phrase, but not all of them. I examined these results (each in the broader context of the page they were printed on—as an item headed “From a Paris Paper” that appeared under a “London” heading almost certainly did not indicate that a North American interacted with a French print) and aggregated them. Second, based on the results I found, I then searched headings for particular newspaper names (as well as a few abbreviations and alternative names), excluding the phrase “from the.”


This data-gathering process produced an enormous number of direct citations in U.S. newspapers, but it cannot be considered definitive, for several reasons. First, America’s Historical Newspapers does not include every eighteenth-century newspaper in (what is now) the United States—though it does include most of them. Second, among the many newspapers that do appear, several have limited or inconsistent indexing of their headings. Finally, Optical Character Recognition (OCR) can be spotty for early modern print, and misses many citations. I suspect that my dataset of Canadian citations contains a somewhat larger proportion of possible citations than the U.S. dataset. As a result, I have avoided making direct comparisons between these data except where I have noted this.

But even if this dataset included every possible direct citation, it would still have several important limitations. First, it does not describe, and does not claim to describe, where all news came from—only news that a printer bothered to cite. In some cases, they likely chose to offer a full citation of a newspaper in order to make a point. Consider, for example, a citation appearing in the Boston Columbian Centinel in 1791, which noted that a news item from the London World was “Extracted from a rank British ministerial print.”[1] Likewise, the Connecticut Courant in 1776 offered an extract “From the London (or lying) Gazette of May 3.”[2] In both cases, a printer emphasized the source of a piece of news in order to suggest that readers should approach it with suspicion. These kinds of citations were not especially frequent, but they do suggest that on some occasions, printers used direct citations in order to draw readers’ attentions to trustworthy or untrustworthy sources. In these circumstances, a direct citation would likely be less-than-perfectly-representative of the paper’s other news sources. Still, this was unusual. Most direct newspaper citations were not intended to put readers on their toes—particularly because so much of the news cited was relatively mundane.

Another note of caution arises from the fact that news reprints were only one of many sources of news circulating around the Atlantic world. The two most important other sources were letters and conversation. Oral communication is impossible to quantify, because it was only occasionally recorded in other media. Letters are scattered across scores of archives, making them difficult to understand in a coherent manner. Letters published in newspapers can be studied, in much the same way that I have attempted to understand news reprints. But the enormous number of these letters, and the difficulty of determining which newspaper first printed a letter (and thus where how that letter travelled) has dissuaded me from attempting to quantify them at this time. Nevertheless, anecdotal evidence and data from Canadian newspapers (which I collected while examining them in microfilm form) suggests that letters moved somewhat more chaotically and unpredictably than newspaper reprints. To some extent, this should not be surprising. Some of the forces that limited and corralled the movement of newspapers—especially commerce, as I discuss below—did not affect letters to nearly the same extent.

But there is reason to think that for a great number of North Americans, newspapers were the most important source of information about the revolutionary Atlantic. News about revolution was often complex and confusing. Newspapers provided the most detailed and comprehensive accounts. In letters from the era, correspondents often referred their readers to the newspapers for detailed accounts of revolutionary events, choosing only to direct their interlocutors’ attention to certain items or to set others in context. Moreover, most North Americans did not receive transatlantic letters or engage in conversation with travelers and mariners. Because the vast majority of the continent’s inhabitants lived outside of port cities, they generally received their news at several removes from maritime arrivals. News from abroad, whether in the form of foreign prints, letters, or rumors, collected in newspapers that were distributed into the continental interior. From there, they set a broader agenda for conversation, epistolarity, and further newspaper reprinting.


This dataset is imperfect, but it provides what I believe to be a reasonably representative picture of newspaper reprints. Rather than focusing on snapshots of particular moments, which may be somewhat skewed, I focus on broad patterns and changes over time. In order to see these broad patterns more clearly, I have generally grouped the data into five year increments.

[1] Columbian Centinel, Sept. 17, 1791.

[2] Connecticut Courant, Aug. 12, 1776.