“The first factor to consider in an authorship question is the number of candidates involved.”
-Hugh Craig, “Stylistic Analysis and Authorship Studies,” in A Companion to Digital Humanities (2004).
The world was my oyster, the path forward clear and bright. Armed with access to the JGAAP tool, “a modular framework for authorship attribution using the object-oriented capacities of the Java programming language” (Juola 2008), and text files of articles that appeared in Home Monthly circa 1896-1897, some of which were published under Willa Cather’s byline and others anonymously, I sat down to generate and triangulate statistical data that might indicate authorship of the anonymous articles. I carefully uploaded my Known Author and Unknown Author texts and ran them through variations of Events (e.g., Character Bigram, Word Length) and Statistical Analyses (e.g., Levenshtein Distance, Histogram Analysis).
Whatever the combination of events and analysis, the results were clear. Curiously, suspiciously clear. Cather emerged victorious in every instance, tagged as the author of every text in every incarnation of analysis. Something was wrong. Brian Pytlik Zillig suggested I add additional Unknown Author texts to see if they, too, were matched with Cather as an author. And so “The Fall of the House of Usher,” “Jabberwocky,” and Mansfield Park were uploaded, analyzed, and returned; Author: Willa Cather.
A befuddled e-mail to JGAAP creator and authorial attribution expert Dr. Patrick Juola solicited a speedy and helpful response. Dr. Juola reassured: “The way JGAAP currently works is to identify the most likely author of a document from among the group of testing authors represented. If there is only one testing author (Cather, in this case), then there is only one candidate for most likely author — Cather is, as it were, running unopposed.” In short, JGAAP, which, unfortunately, at this stage lacks an instruction manual, requires that users submit queries that include more than one possible author for any set of Unknown Author texts. This requirement, as the epigraph from Hugh Craig that opens this blog post suggests, is not simply a quirk of the tool but is, rather, a tenet of considered authorial attribution analysis.
The problem, of course, was that, for our purposes, namely identifying authorship of discrete short texts in the close context of a publication, using Lewis Carroll as distractor author would simply not work. Optimally, the authorship question would be posed around identifying the true author of a text from a list of real potential authors, whose styles and patterns were likely more closely aligned than Cather’s and Poe’s. In need of a suitable distractor author, I sought out Andy Jewell, the aforementioned Cather and kolache expert. Andy provided me with important research into the authorship of the disputed Home Monthly articles and pointed me towards scans of the Monthly, from which I might select additional texts for comparison.
As Timothy W. Bintrim details in his 2004 dissertation, “Recovering the Extra-Literary: The Pittsburgh Writings of Willa Cather,” Cather served as managing editor of the Pittsburgh-based Home Monthly from August 1896 to June 1897, overseeing eleven issues and contributing many articles, including pseudonymous and anonymous writings. Having arrived in Pittsburgh as a twenty-two year old to take the reigns at the Home Monthly, she would remain in the area for a decade. As Bintrim reveals, Cather was notoriously protective of her work once she had achieved success as a novelist, becoming: “ruthless in censoring the fiction, poetry, and journalism of her first career, invoking the power of copyright to keep most of her extra-literary writing out of print during her lifetime.” The scholarly attempt to identify and make available the writings Cather produced as a twenty-two year old editor of and contributor to Home Monthly works against Cather’s own “resolve to have her first career buried with her.”
Based on Bintrim’s work and an examination of the contents of Home Monthly, I identified two potential approaches to the problem of the distractor author. The first is simply to select authored pieces in the monthly that might be uploaded into JGAAP as comparison texts, tagged with their authors. While it may be unlikely, for example, that Erasmus Wilson, who regularly contributed a column on boys and girls to the Monthly, penned any of the short pieces we mean to examine as Unknown Author texts, his columns will nevertheless present a contemporaneous authorial style present within the same publication. A second approach would be far more discrete: Bintrim writes of a scholar who has attributed pieces in the Monthly appearing under the byline “Mildred Beardslee” to Cather. Bintrim rigorously disputes this attribution, with the support of extensive textual evidence. We might weigh in on this smaller Beardslee/Cather with the support of JGAAP, though this again incurs thorny single author issues.
While wondering at this problem, I have been using OmniPage software to perform optical character recognition on TIFF images of Home Monthly, scanning and correcting articles that might be useful. Despite certain setbacks in my framing of the Cather attribution problem for ingestion into JGAAP, the week also offered the opportunity to recast the problem in the context of ongoing attempts by Cather scholars to establish authorship of Home Monthly texts and to flesh out Cather’s involvement with the publication. Bintrim quotes a scholar, writing in 1950, who predicted that identifying Cather’s unsigned writings from her years in Pittsburgh would “’offer someone an interesting, if vigorous exercise in literary detection.’” Almost sixty years later, it is exciting to contribute to this ongoing attribution effort, and to witness the array of techniques that have been deployed in pursuit of definitively linking Cather to early pieces that provide unknown biographical details and “contain clues to her friends, benefactors, and antagonists in Pittsburgh” (Bintrim).
(Image: “Group 6 Women, Willa Cather 2nd from right.” Held at the Willa Cather Pioneer Memorial and Educational Foundation. Available online through the Cather Archive.)
