research, DH, thank a feminist

For the past year I have been working with the journal Signs on their fortieth anniversary project. Signs@40: Feminist Scholarship through Four Decades is now live. It features an interactive visualization of a topic model of the journal’s archive that I created in collaboration with Andy Mazzaschi, Susana Galán, Laura Lovin, and Lindsey Whitmore. The model is accompanied by commentary—both our own guide to exploration and commentaries by Signs editors and contributors past and future. The Signs group also created a series of thematic tables of contents for the journal and adapted a co-citation network visualization originally developed by Jonathan Goodwin. The whole provides many pathways for exploring the work of this central institution of feminist scholarship.

We will shortly make the source code available on github, where I’ll also add a few more technical notes on our modeling process and our visualization design. Here are some early thoughts about the site itself, now that it is at last whole and public:

For me the commentaries are particularly gratifying, since they actually use the model to make arguments about tendencies in the archive. The past, present, and future editors of the journal explore the visualization, deeming some of the apparent trends it reveals interesting and meaningful, and pointing out some of the further directions in which it would be necessary to take the analysis. For example, the incoming editor, Suzanna Danuta Walters, argues that the visualization provokes questions about what is happening outside the discourse of this single journal, where the “decline” of certain themes may really be their migration into other venues. Inside and outside, too, changes may be driven by institutional forces. This is a good caution against imagining the model as a totality, rather than itself partial evidence; nonetheless, Walters is willing, despite her skepticism about “the digital” as such, to consider this exercise as a way of discovering patterns in the field that need now to be explained. In the same vein, in her comment Brittney Cooper considers the model’s indicators of scholarship on Black feminism and finds it both potentially useful for making connections and notably incomplete when it comes to the key category of intersectionality; in the model we constructed, this term is folded into other, less specific rubrics.1 The irony is notable, but Cooper generously frames it as a direction for further work, as indeed it is. Many of these questions seem to necessitate moving from unsupervised modeling techniques in the exploratory phase to some kind of supervised modeling in the explanatory phase of analysis.

Dana Britton examines a topic that particularly struck me, one we labeled Quantitative methods. Britton argues that the apparent decline in quantitatively-oriented social science in Signs is real, and partly an effect of the appearance of new field-specific venues for studies of women and gender. Feminist quantitifiers in sociology or economics are, she hypothesizes, now more likely to turn to specialist journals in their disciplines. For a sometime quantifier in literary studies like me, these arguments are particularly intriguing. Personally I find the interdisciplinary ethos of the early years of Signs striking as a model of how qualitative and quantitative forms of cultural analysis might join in a single discourse: the “new scholarship about women,” the founding editorial called it, with a mission to foster “the accurate understanding of men and women, of sex and gender, of large patterns of human behavior, institutions, ideologies, and art” (v). The italics are mine: the importance of the “large pattern” seems to me to provide one justification for the founding editors’ insistence on multiple methodologies.

I believe in the significance of large patterns, and I believe that feminist scholarship teaches how to grasp significant intersections of large social-cultural patterns with individual experience. But I do not mean to claim that we can now grasp the total pattern of Signs, especially on the basis of something as simplistic as a topic model. We have tried, in our interpretive apparatus, to emphasize the openness and noisiness of the model: it provides useful contexts for thinking about women’s and gender studies as represented in Signs over time, but it cannot provide definitive maps. And the best antidote to totalization is the actual writing of the journal. Your explorations in Signs@40 will tend to feel incomplete until you have made your way to the writing in the journal itself and begun to (re)read. Nonetheless, the pattern-finding possibilities of the visualization may change the way you read.

The relative decline over time in the share of quantitative methods in the journal’s pages, however, doesn’t mean social science has exited the journal: quite the contrary, its centrality is such that Catharine Stimpson, the founding editor, who works in literary studies, sees topic-model evidence of the humanities’ decline. (Literary-critical topics are more concentrated in the journal’s first two decades.) So perhaps, as Britton and Stimpson both suggest, it is better to consider a broader conception of interdisciplinarity: might it even be that the topic model indicates for us the gradual decline of certain disciplinary discursive distinctions in the language of the journal? To generalize reliably about this lies beyond my expertise, since, though I consider myself a feminist, my departmental home is English studies and my subfield is twentieth-century literature, so I have some kind of ethical duty not to mansplain the history of women’s studies to you.

This final remark, then, is rather speculative, but it comes from my sense of the valuable example Signs offers for scholars across the human sciences interested in the intersection of disciplines. Here is one case. I went back to Sandy Alexandre’s 2011 essay “From the Same Tree: Gender and Iconography in Representations of Violence in Beloved.” “From the Same Tree” is compelling literary criticism, but its predominant topics in our model are Black feminism, The social, Women’s roles, and Historical terms: thus the model provides a reader with an imprecise but not inaccurate signal of the interdisciplinarity (and intersectionality) of Alexandre’s approach, combining literary analysis, cultural history, and visual culture studies. At the same time, Alexandre’s essay shares the issue with a symposium on Love Exiles, in which two contributions use quantitative data and two do not. This sort of methodological coexistence is the norm for Signs, and it is one useful feature of the topic model that it allows you to consider this across the run of the journal, simply by noting the way topics that are methodologically marked share space in a given year.2 One thing that makes Signs special, then, might be its position as a model—a changing model—for the production of knowledge across disciplines. But I might like to put it more strongly. Have you wanted to combine quantitative and qualitative methods in a new synthesis to study large patterns in culture and society lately? Then join me in thanking a feminist.

  1. Software limitations prevent us from making it possible to display the topic breakdown of all 62,522 word types included for modeling, and included only words ranked 50 or above in some topic. The word intersectionality is allocated by the model in 736 of its 743 occurrences to the topic we have labeled Politics of difference, which, as Cooper notes, seems to have encompassed the discourse of intersectionality. It ranks as the 52nd most prominent word.
  2. Zooming in on the streamgraph of topics over time helps; once you have scrolled to zoom, you can hold Shift and drag to pan.