December 1, 2015 WORD LAB – Beth Seltzer

Beth Seltzer is a library postdoc at Penn Libraries, who works with Victorian detective fiction and is a recent Temple University Grad. You can find her on Twitter: @beth_seltzer. Her title is ‘Drood Resuscitated’!

The presentation is on her work on Charles Dickens’s last novel, The Mystery of Edwin Drood, which was serialized until and just a little after his death because he had written a few more installments before he died. Many unanswered questions at the end of the book, including about the characters and whether it was even going to be a detective novel in the first place.

Reader response: Fictional continuations, literary criticism, spiritualist attempts to get in touch with the ghost of Dickens! There are about 400 texts 1870-1939 in response to Edwin Drood; a “shift” in Dickens studies (although quite arbitrary) in 1939 helped Beth set her cutoff at that date. It’s an uneven corpus by year, dying down over time and then coming up every so often (with a burst around 1904, because of The Dickensian journal as a platform for Drood criticism).

Method 1: “Manual Tracking” where Beth read through it all manually and kept track of notes in a spreadsheet about various attributes of the criticism/responses, such as about plot points and style of the novel.

Method 2: “Topic Modeling” to find clusters of related words in the texts. Beth collected all available texts, using cleaned Gutenberg texts when possible but it wasn’t a very Gutenberg-friendly corpus; most were run through ABBYY FineReader and cleaned with regular expressions or simply retyped, with about 95% accuracy overall. She got 20 topics, run multiple times and tried excluding character names.

Then there’s the “weird stuff” – for example, the “Sprit-Pen” version of the novel supposedly written by Dickens through a medium after his death, and a 1927 seance with Sir Arthur Conan Doyle through a medium! There was also a 1914 “mock trial” by authors and critics, putting John Jasper on trial for the murder of Edwin Drood.


Over time, the belief that the novel was predictable and is conclusive drops off drastically. Moreover, speculation grows over time – “a shift from the author to the reader.” What does that mean?

In the topic models, there seem to be more author-centric “topics” at the beginning, with a focus on Dickens, other writers, his other novels, and so on. Later, you see reader-centric topics more focused on theorizations of the novel, critics of the novel, problems in the book that readers are engaged with. Comparing the two topics, you see that theorizations starts low and goes up over time, whereas the Dickens-related topics decline over time.

Readers also become increasingly creative and autonomous over time, feeling like they can make more authoritative statements about their theories. You see an increasing range of characters over time being suggested for the “real” identity of the Datchery character, including more speculative ideas such as a new character, multiple people, or “Datchery (not in disguise).”

Beth also did some network analysis, using an annotated bibliography by Don Richard Cox. She set up her analysis of the citations in Gephi, sized based on how many other texts refer to them, colored according to genre. The graph revealed that the biggest pieces of conversation were early literary criticism, and the readers were interested in themselves as readers – rather than talking mostly about specific editions of the text, for example.

The readers also appropriated new discourses to talk about the novel. People were interested in the “genius author” idea in other works of Dickens, the mock trial mentioned before (which also inspired imitators, and people writing about it/them), and also “Rochester,” the real city that the novel is modeled on. The topic of Rochester, revealing a bump of literary tourism at the time, was something that Beth hadn’t picked up on in her manual reading of the texts, so it is a point where the computational analysis suggested something new and unexpected.

Summing up:

Readers increasingly push the boundaries of the text and cast themselves as co-creators of the text.

Questions and Brainstorming!

What role does Edwin Drood play in our society now and what does it tell us about readership? Move forward chronologically and look toward the contemporary moment.

Popular vs. scholarly audience and attitudes toward the novel: Scholarship says it is against speculating about the ending now, as opposed to in the early 20th century. YET, many critics speculate in those very same works! Perhaps we are inevitably drawn to these questions and just can’t help ourselves.

How do you responsibly handle the massive amount of Drood stuff out there now? What about the premise that this helps us to understand readership and/or detective fiction – does this hold up? What about the data that gets left out (including criteria that Beth didn’t track), and what are other tools and techniques that can help us understand the corpus we are working with?

Our WORD LAB discussion…

Beth notes that while a spoof was being published simultaneously with Drood while it was being serialized, not much in terms of criticism of it while it was ongoing.

What are the differences/continuities between criticism of finished versus unfinished novels? Or how about between criticism of other unfinished novels? Beth notes that there wasn’t the same type of draw with other unfinished novels, and speculates that it might have something to do with Drood being a mystery. The question is, is there something about how “firm” the ending is that makes it so amenable to speculation?

Is there a point in the plot where the feeling of “why didn’t you tell me it doesn’t have an ending?” becomes hightened? Perhaps there’s some work to be done with looking at the text itself and analyzing its structure. Indeed, no one perceives it as a complete story, so that’s a sign that there’s something that makes people invested and really care how it “should” have ended or what “really” happened beyond the ending. Is there some kind of pattern that gets disrupted at some point that makes people interested/invested?

Dickens may not have been an object of study in the early 20th century, and a lot of the criticism is in The Dickensian which thinks it’s a scholarly journal but does have a fan culture around it; many of the writers were in the “public intellectual” realm than purely academic conversation. Was the conversation more popular? And yet the “popular conversation” is focusing on a critical conversation, and at the same time an ongoing conversation about a controversy. If you go forward in time it would be interesting to see whether there’s an increasing split between academic and popular discourse.

Labeling the pieces of criticism geographically would also be interesting – to see if some approaches are more common in certain places and if that maps across time. But at the same time, they’re only in the US and UK.

The reaction over time (just making a generalization) seems to have been speculation in the 19th century, but now, more answering the question “did you like it?”. “Is it good or not?” not “what happened?”.

How do you get at reader response at all? For the Victorian period, you get very painstaking studies of diaries, letters, etc., looking at psychology of reading in the 19th century, and so on, and yet this reader feels very elusive. Is there any way to get at these readers, or is there a point to even trying, especially if it can’t be done?

How about seeing more correlations, like the comparison of fiction thinking Edwin died, and fiction thinking he didn’t? There seem to be a lot of possibilities for exploring things like this. For example, plotting topic models across all the 12 variables Beth has rather than over time. Plot by “belief in deadness,” etc. It would make a lot of graphs!

How about comparison to a similar body of texts – such as a number of reviews or discussions of Dickens from the same years? There’s also the Dickensian journal with lots of articles for comparison.

Track evidence that’s being marshaled for arguments – although how would you do that, or what would the categories be? In the medium stuff you’re supporting your argument with “direct access” to the author, but with the later pieces, are they doing close reading, doing something tongue-in-cheek, or just speculating, or something else? It might give more detail to the shift from author to reader. Beth was actually tracking this, but then gave up on it because it felt like her calls about what the argument support “was” had gotten very subjective.

Leave a Reply

Your email address will not be published. Required fields are marked *