“A conversation with data” – Dan Cohen on close and distant reading
Recently, Dan Cohen posted the pre-print to an article on his blog that struck a chord with me, and on which I would comment on here. (Dan Cohen’s blog, by the way, is currently the “featured blog” here.) The article, co-written with Fred Gibbs, treats an issue Toma and me touched upon briefly last week in the comment section to another post here: the (mis)perception of the digital humanities as an exclusively quantitative approach to cultural artefacts, and the way digital humanities research actually may work best when it moves back and forth between “traditional close reading” and “computer-enhanced distant reading”, between patterns and details – and, I may add, between the birds eye’s view, the dragonfly’s gaze, and microscopic vision. Combining these approaches leads to an engagement with literary texts which Dan Cohen amd Fred Gibb beautifully characterize as “a conversation with data”.
Rather than just asserting this, the authors then go on to discuss several projects in which they used this kind of combined approach. In the first project, they looked at the use of a large number of keywords in the titles of more than 1.6 Mio books published during the “long nineteenth century” to which the Google Book project gave them access. They investigated the correlation of such terms’ relative frequencies with tendencies in the history of ideas elaborated in a classical “pen-and-paper” monograph (Walter E. Houghton’s The Victorian Frame of Mind, 1957). They found a sharp decline of words like “God” and “Christian” in book titles, setting in as early as just after 1850. This is earlier than the start usually assumed for the crisis in faith of the nineteenth century.
Another study was conducted in the context of the “Criminal Intent” investigations into the “Old Bailey Online” database of court proceedings (which Robert Shoemaker nicely showcased in Paris, recently). The Old Bailey proceedings show a rise in court cases involving female bigamy, and the authors wanted to find out if other written evidence could give clues on why this should be the case. One line of argument involved the changing perception of marriage in the Victorian Age, and this was analysed through an investigation of two-word sequences in the Google Book corpus where marriage was the second word. Some telling combinations rose in frequency, over the course of the nineteenth century (like “hasty marriage” or “loveless marriage”, for example), while others fell.
So, what did this research make possible that would not have been possible without digital tools, and how did it combine close and distant reading? In the first project, testing hypotheses from a traditional monograph by the evidence from 1.6 Mio book titles certainly gave added insight and nuance to the history of ideas in its relation to the publishing industry. The combination of close and distant reading was in combining existant issues or themes defined by close reading with new answers arrived at with methods of distant reading. In the second study, I am slightly less impressed by the evidence from the Google Book search (at least with the one presented in the brief article). Also, the close reading aspect of the approach seems to lie more in the fact that from the bigram results, one could have dived back into close reading of individual books, rather than the actual practice of doing so.
Still, this is cool stuff, and certainly supports the authors’ claim that this kind of research “help[s] flesh out prior historiography as well as provide new perspectives that advance historical interpretation”. What we all need now, of course, is access to the same kind of advanced Google Book data which the authors had access to!