Close-Reading the Database

Lately, in my academic reading, I keep encountering a peculiar mode of textual argument, one that build associatively rather than logically, as if there were a transitive property of meaning. Sometimes, connections made within the logic of one set of texts (paper and lambskin, jewels and water) are chained to those within a different set of texts (lambs and meat, water and blood). A means B means C.  Other times, in a sort of Reader Response writ large, we learn that an early modern reader “would have associated” X with Y and from there with Z. The connecting threads are woven of simile, historical coincidence and conjecture, and suggestive phrases like “Perhaps it is no accident,” “it would not be too far-fetched to imagine,” “similarly,” “much as,” and “network of associations.” The resulting arguments are often incredibly learned, offering rich accounts of their conceptual landmarks, even as they are knotted together largely by the accumulation of puns and coincidences.

Sailing to sea in a sieve

Sailing to sea in a sieve.

Such projects close-read the world, applying the associative tracking that allowed formalist critics to offer an account of the meaning of image clusters within a poem to working out the resonances of a particular object or emotion in the world more generally. In so doing, they assume that meaning really is a closed and singular network, that once we’ve paid our fare, we’re free to travel from station to station as long as we’d like until we arrive at a destination we find suitable. At their best, such circuitous routes complicate our mental maps of the text, revealing new connections and different approaches. At worst, they seem (like automobile GPSes) to follow an imaginary map that never quite corresponds, sending us down disused highways and interstates that were never quite built.

The ways authors describe and defend this mode of reading are fascinating. I’ve recently seen it framed as a sort of object-oriented philology (bromides about the materiality of early modern language enabling basically sociological accounts of meaning), as inventive source study, or as close-reading at a distance. We could add other touchstones: New Historicism, of course, both in using the particular minor detail to stand in for the whole structure and in its willingness to read the canonical text through details that were largely extrinsic to it. Queer theory, in its use of puns to uncover desires that have been repressed. The way the new rhetorical criticism’s focus on figures as structures of thought allowed it to travely freely between instances.

But it seems clear to me that this mode of reading has emerged in larger part from the explosion of digital tools. When we try to close-read the world, we in fact are close-reading the scholarly databases and institutions with which we try to understand that world. Such tools are incredibly useful, allowing us to clear away the underbrush of received information by digging through all the early modern instances of a construction, a concept, a figure, or a phrase. We can trawl the DNB for unexpected connections between individuals or instituitions, dig through the Old Bailey trials for anecdotes, use CQPWeb to follow a grammatical construction, and move immediately from all these searches to clusters of scholarship in the MLA bibliography and back again. History makes itself visible to us as a combination of failed searches and unexpected connections.

My point is not that these types of inquiry are novel. They’re not: a sufficiently ingenious scholar, with sufficient resources, could have done any of them in 1920. But I do some number of these things every single day, and the result is an overgrowth of associations that kudzu-like sometimes conceals the underlying structures. As a process, searching databases always yields unexpected connections, simply because one keeps searching until one finds one.

The inevitability of this process makes me wonder how to judge the arguments that emerge. How do we assess arguments that by design are associative, trying both to interrogate deep patterns in a culture and to trace out some patterns of their own? How much weight are we willing to put on the metaphor of a network of associations? A net may sometimes catch a fish but will never hold water. And how do we pay sufficient attention to the ways that the limits of our tools constrain what we find?

I have long been moderately skeptical of Big Data approaches to the humanities, all too aware of the ways that our algorithms can distort our inquiries and that we, in turn, misapply the results we get to the problem we want to solve. But those of us in the land of Small or Medium Data may well need to think through the same challenges.



Image: the Witches from Macbeth, by an unknown Edwardian or Victorian artist. From the Folger Shakespeare Library.

I am grateful to Laura Kolb for taking a look at an earlier version of this post.