Proposing my digital humanities dissertation

Spurred on by some of my colleagues as well as a virtual visit I just had with current students at my PhD alma mater, I’m finally writing up my experience proposing my digital humanities dissertation, including sharing the proposal document itself from fall 2014, for a dissertation that I defended in spring 2016.

This was my experience, which I offer up as one possible model for a DH dissertation. Expectations around proposals in particular can vary heavily not only between fields, but between individual departments, so your mileage might vary. To put some big-picture qualifiers out there at the start:

1. My research was heavily inclined towards what is increasingly called cultural analytics i.e. computational analysis of humanities data to drive arguments about cultural history. My work is just one slice of the possible modes of digital humanities research, and of formats of DH dissertation. I’ll discuss the ramifications of this for my own proposal and writing process.
2. I was based in a decently-ranked but non-Ivy art history department at a public university in the U.S., with an advisor who was also a prominent art museum curator. I believe both of these facts contributed to my success in getting this proposal and dissertation accepted, where I may have run in to more resistance at a more elite program.
3. As I’ll note in the discussion about my work schedule, I had a generous fellowship structure, with dedicated research and travel funding, that also set me up for success.

The format

After a few exploratory talks with my supportive advisor about my topic and approach, he nervously asked, “so… this dissertation, is it going to be… digital?”

I assured him that, although the research would involve a lot of coding and computation, the result would be a very recognizable format: a beefy PDF he could print out and mark up in pencil. There would just be quite a few charts in it alongside images of artworks. Amanda Visconti has an excellent walkthrough of different considerations about the format of your dissertation, and many good arguments for why you might think about modes other than a big chunk of written text. Producing something legible as art history to my committee was one reason for going this traditional route. But additionally, it turns out that a linear written argument is a pretty useful format when you need to step from explaining a complex (and new-to-your-audience) quantitative method, to then presenting and interpreting the results through in-detail case studies.

Linear doesn’t have to mean dry. My advisor and I made a deal that this could work out as long as I was able to explain all the analyses in as plain and relatable prose as possible, even if that meant creating some pretty lengthy and metaphor-filled methods sections in each chapter. We also agreed on following a shared pattern for each chapter: every 20,000-foot view quantitative analysis had to be accompanied by stories about individual artists or works that translated the implications of big-picture arguments into a familiar scale for my readers. This is necessary for pretty much all cultural analytics work.

Expectations for a proposal

Even within our department, expectations for the scope of a proposal could vary a lot depending on your advisor’s approach to the process. Generally speaking, though, I got the impression that 80 pages was way too much, but the 20-30 pages of a term paper was not quite enough.

Within that rage, most proposals needed to share the same basic components:

1. Concise statement of your research goals
2. Articulation of your contribution to existing scholarship
3. Breakdown of the structure of the dissertation chapter by chapter
4. Proposed timeline for research activities as well as when you’d be handing in chapter drafts

The elevator pitch

The sheer quantity and variety of artistic prints from the seventeenth-century Netherlands, and the number of individual printmakers and publishers involved, challenge traditional models of art historical argumentation based in case studies. I wanted to address this issue by using quantitative methods, particularly network analysis, to analyze large-scale changes in the organizational patterns and artistic strategies of reproductive printmakers and publishers in the early modern Low Countries.

Contribution to the literature

My research goals were twofold:

1. to reexamine several commonly-held historical narratives about artistic print production in the period
2. to demonstrate that computation - network analysis, in my case - with collections data had a lot to offer art historians, particularly those who study printmaking

Therefore, articulating the place of my proposed research in existing scholarship meant both grounding it in art historical literature as well as historical network analysis from a variety of fields. If DH approaches feature in any significant way in your dissertation, it is a good idea to make sure you can situate your work within the larger DH literature. It’s not a new field anymore, and you are certainly not the first person to think about digital approaches in your own discipline! Depending on how central those approaches are to your research, this “situating” work may be brief, or quite deep.

This will be helpful to any readers on your committee to whom DH is new (and possibly suspect): demonstrating that it has pretty deep scholarly roots by now, and has been successfully applied to disciplinary research questions can go a long way towards addressing good-faith questions about the viability of data-driven methods for your proposed work. On the other hand, if you have a committee member or two that do know something about the field, they’ll rightfully expect you to be able to do that kind of lit review work as well.

Chapter breakdown

I designed my dissertation as a cluster of three data-driven experiments working from a set of shared data sources. I won’t re-summarize each chapter here - just look at the actual proposal or the intro to the diss if you want that. But I think two key moves set me up for success here:

First, I structured each chapter’s experiments such that they’d be useful no matter what results I got.

I don’t think it cheapens the dissertation experience to acknowledge that risk management thinking is a useful tool in the process. What happens if I go to this archive and don’t find the documents I hope to see? What happens if my sources or data tell a very different story from the one I thought they would? How do I manage the very real risks of maintaining my funding and juggling the deadlines at the department, college, and university level?

In my case, I framed my quantitative questions as “either/or” questions in a very broad sense: Did the Netherlands become more centralized later in the seventeenth century? Or less? Did Dutch printmakers overall become more intertwined with international printmaking partnerships after the revolt, or more inward facing?

Of course the complete answers to these questions were not simple binaries, but using these broad strokes helped me establish the stakes. It clarified the big implications for our historical narratives based on what my data-driven analysis turned up. As I’ve argued before, there would be historically-plausible explanations for all of these questions whether they turned out (based on my particular set of evidence) to be “yes” or “no”. This meant it was all the more vital to actually run these experiments!

Articulating that my questions were important in and of themselves, no matter the answers I found during my research, was a necessary task from a scholarly standpoint. But in a very practical sense, it also insulated me from the danger that my data might insist on a different story than the one I had hypothesized. I wouldn’t end up trapped with a dysfunctional chapter if my sources pointed in a different direction than I had expected.

Second, at least one of my non-lit-review chapters was just about “done” by the time I finished this proposal.

My impression from dissertators in my department was that it was invaluable to base your dissertation around a strong term paper, an MA thesis chapter, or some kind of already-finished work. Starting entirely anew - without existing in-depth knowledge of the literature in your specific focus - could add a whole semester or more onto your dissertation.

Although I’d focused on one printmaker for my MA thesis, I didn’t have an extant course paper doing any of this network analysis stuff when I finished my PhD exams and started the process of proposal writing. Therefore, I took the maximum-allowed time between completing my comprehensive exams in spring of 2014 and defending my proposal December of that year. During that time, I was able to build on exploratory work during one of my graduate assistantships to really solidify my understanding of museum data and historical network analysis, and do enough concrete experimentation to determine what the appropriate research questions and scope of my dissertation would be. During that time, I acquired all my data, I had figured out exactly which network measures I was going to use, and I taught myself enough R to have a very good sense of the results of my first experiment.

A lot of this work sounds like I was already just writing the dissertation, and of course I was. But it gave me the time to identify dead-ends vs. productive pathways before committing to them on paper in a proposal. There was no benefit to trying to propose earlier than I did, and I really needed all of that time to fully conceptualize and articulate the work I was going to do.

My work schedule and outside reader

My proposal also included a work schedule, where I essentially laid out a schedule of 1-chapter-per-semester in order to graduate in December of 2016, a spacious 24 months after defending my proposal. I ended up needing less time than I projected, which is not at all the case for many dissertations.

This is where my funding setup at UMD worked out very well: for fall 2014 / spring 2015 I was embedded in a curatorial fellowship at the National Gallery of Art. This greatly eased my access to the excellent research library there, which certainly did not hurt the writing of my proposal. I also had a full fellowship year to support me through fall 2015 and spring 2016, which meant a full year to focus entirely on researching and writing.

In addition to the funding, I also attribute some of this speed to my own advisor. As a museum curator used to working with everything from scientific examinations done during conservation, to borrowing centuries-old ice skates from Dutch enthusiasts, he had a pretty capacious sense of what could be used as evidence in the pursuit of learning more about artworks. He also had no reservations about his students going into careers in museums, foundations, auction houses, or galleries rather than pursuing a faculty job. Even when a research topic wasn’t his exact thing - mine certainly was outside his wheelhouse - he usually did not stand in the way when he could see that the rest of the community was waiting for someone to graduate.

It was also in this section that I asked the committee to accept something fairly radical: that most of the time I budgeted for research would be spent curating collections data and refining my R skills, not traveling to archives or museums. Being able to draw heavily on a dataset already compiled by others was key to this dissertation. I absolutely would not have been able to learn the computational skills necessary had I also needed to construct my dataset from scratch. To be honest, I’m not 100% convinced my committee grokked that part of my proposal when they signed off on it, but at the end of the day it all worked out?

Finally, I included a list of potential outside readers from UMD’s school of Information Science who could bring expertise in network analysis to my dissertation. There’s now a pretty solid network of digital humanities and cultural analytics experts who can serve this role on dissertations. I’ve even had the pleasure of serving as an outside reader in the past year for another digital art history diss. As with the job market, early networking through presenting your work at conferences and even just at interdisciplinary events on campus (which is how I made my connections) can help you out a lot here.

The results

Everyone was happy enough, even/especially my outside reader, and I had my dissertation accepted without revisions.¹ The official dissertation deposit still seems to be embargoed, but I put up a copy on my personal site if you really want to read it. I never envisioned the dissertation as a book, and although it could probably have been shoehorned into one if I had decided to go down the research faculty track, I’m glad that the chapters have instead come out independently.

An expanded version of Chapter 1 - the lit review - ended up being the last to go public, finding a perfect home as “Tangled Metaphors: Network Thinking and Network Analysis in the History of Art” in The Routledge Companion to Digital Humanities and Art History.

Chapter 2, exploring centralization and decentralization in printmaking, was the first to come out, in the form of an article in The Journal of Digital Art History. I submitted this chapter before I defended my dissertation. The reviews were very influential in helping me reframe what some of my analyses were doing, in ways beneficial to the dissertation itself.

Chapter 3 swapped places with chapter 4 in the final dissertation since it continued more naturally from Chapter 2. This work analyzed the international collaboration patterns of Netherlandish printmakers. I think this is the most successful and exciting result of this dissertation, and appeared alongside a variety of other seminal art historical network analyses in a special issue of Artl@s Bulletin. I submitted it several months after I defended.

Chapter 4 - the genre diversity experiment - is the black sheep of my dissertation. I think the disciplinary research question was right, but I was never fully satisfied with the computational measures I used to address the question, particularly given the incommensurability of the artistic print datasets to the paintings dataset I used. That said, it did the work it needed to do for the dissertation by looping the subject matter of the prints into what was otherwise a genre-agnostic project. I ended up sharing this as a conference paper at DH2016 in Kraków but didn’t pursue it for further publication.

Other posts related to my dissertation