What was advertised in a colonial American newspaper 250 years ago today?
“SOLOMON SOLOMONS … A fmall Affortment of JEWELERY.”
Earlier this week NPR commentator Cokie Roberts caused quite a hullabaloo when she suggested that historians had significantly inflated the frequency of advertisements for abortion providers that appeared in nineteenth-century newspapers. Roberts stated, “There are many articles by abortion rights proponents who claim the procedure was so common that newspapers advertised providers. Look, I did a search of nineteenth-century newspapers and couldn’t find them.” Historians quickly responded via Twitter, with Dr. Lauren MacIvor Thompson, a specialist in the history of medicine, public health, and the law, in the forefront with a tweet thread that corrected the record.
In addition to addressing content, historians representing various other fields within the discipline addressed the flaws in Roberts’s methodology. Roberts, a pundit rather than a trained historian, apparently did not realize that the absence of results generated by keyword searches does not mean that the historical evidence was not there. Like many of my colleagues, I pointed out two relevant issues. Both are so fundamental that I discuss them with undergraduate students in introductory and upper-level history courses on the first day that we begin working with databases of historical newspapers.
First, keyword searches have many shortcomings, especially because OCR (optical character recognition) is so imperfect. I explain to my students that computers are often, for lack of a better word, stupid. They do not always recognize or make sense of visual images (photographs or digital scans of historical sources) as effectively as people do. Computer software lacks the necessary creativity and flexibility. This is especially true when working with eighteenth-century printed sources that use the long “s” that looks like an “f” to twenty-first-century eyes. What human readers recognize as “Assortment,” for instance, looks like “Affortment” via OCR. (See Solomon Solomons’s advertisement from the June 7, 1769, edition of the Georgia Gazette.) Yet the long “s” is not the only pitfall for OCR. If the original printed words were not clear or subsequent remediation (photographs, microfilm, digital scans) was poorly done, then OCR has no chance of decoding the words on the page.
Second, when doing historical research it is necessary to think like the people from the period, especially to use the words they would have used rather than impose modern terminology. Context matters. Roberts, lacking an historian’s understanding of the period she investigated, apparently did not choose her keywords carefully or appropriately. When I train my undergraduate students to serve as guest curators for the Slavery Adverts 250 Project, I underscore that they must look for words beyond just “slave” (and that they have to examine every advertisement because keyword searches will exclude MANY advertisements that belong in the project). In addition to “slave,” they must also keep their eyes open for “negro,” “mulatto,” and “wench,” some of the words most often used to describe enslaved men, women, and children even when the word “slave” did not appear in an advertisement.
To demonstrate the shortcomings of keyword searches, I like to provide a practical example of an advertisement that I know exists yet a keyword search will not produce. Consider an advertisement from the June 7, 1769, edition of the Georgia Gazette. I previously downloaded the entire issue via Readex’s America’s Historical Newspapers database. I know that it contains an advertisement for jewelry placed by Solomon Solomons. Say that I want to know the extent of Solomons’s advertising campaign in 1769. Doing a keyword search with his last name would be a good place to start. To make the search as efficient as possible, I set several parameters. I limit the date under consideration to 1769. I restrict the newspapers to be searched to the Georgia Gazette. Then I enter “Solomons” as the keyword. This yields only two results: Solomons’s advertisement when it appeared in the May 24 and May 31 editions, but not the June 7 edition. This certainly tells me more about the frequency that Solomons advertised, but it did not yield an advertisement that I already knew existed! The digitized image of the advertisement is fairly clear (especially compared to many others), yet it appears that just enough ink bled through from the other side of the page to trick the OCR into overlooking this advertisement when doing a keyword search for “Solomons.”
This particular instance is not as “fraught,” to invoke Roberts’s term, as advertisements placed by abortion providers in the nineteenth century, but it is a practical example of how technology cannot substitute for historical expertise and appropriate methodologies for conducting research with primary sources. As many other historians have done in recent days, I encourage reporters and pundits to call on trained historians rather than make misleading assertions based on incomplete understandings of the past and shoddy research methods.
 This quotation comes from Thompson’s tweet thread. NPR has updated the original audio and transcript to excise Robert’s incorrect and misleading assertions.