Mixed Methods
Within the humanities, there is an undercurrent of hesitation regarding its “digital humanities” subfields. The application of digital techniques to literary studies, for instance, has been sometimes depicted as turning the field into a “‘counting’ discipline”, prompting the “hoary binarism of quantitative versus qualitative” (James English, quoted in de Sá Pereira, 2019). Continuing with literary studies, other scholars advocate for the field to ally itself to the social science by becoming more quantitatively oriented (de Sá Pereira, 2019).
Overcoming the qualitative vs. quantitative dichotomy includes the use of mixed methods, widely accepted in the social and health sciences. Mixing of methods is sometimes an iterative process wherein both qualitative and quantitative methods are applied in a single study. It is iterative because different paradigms – qualitative or quantitative – may be applied at different stages of the research. For instance, geography has integrated the use of computationally intensive geographic information systems (GIS) and satellite imagery with a critical theory, combining the approaches of GIS researchers and human geographers in a field of study known as Critical GIS. Such an approach is proposed for the humanities. Methods specific to the digital humanities, spatial analysis enabled by GIS, can be combined with close reading.
Literary pattern recognition fuses digital techniques with more traditional humanistic approaches. This approach was demonstrated on the study of English language Haiku poetry as representative of Modernist text. The goal is to identify the haiku as a distinct literary pattern. Three different techniques were applied to the poems:
- traditional close reading;
- analyzing the poems through a socio-historical lens; investigating how this originally Japanese literary form was historically adopted into an English-language environment as well the social context of this adoption; and
- applying machine learning techniques to them.
Each critical method produced different conceptions of the haiku, with different patterns being recognized. However, a synthesis can emerge when these varied patterns are viewed as complementary instead of mutually exclusive, wherein the limitations inherent in each method is supplemented by the other methods. The different “discrete ontological visions of the haiku”, when synthesized, “…produce a more comprehensive picture of the English haiku as a social or cultural milieu—part of a broader Orientalist style that circulated in the early twentieth century” (Long & So, 2016) (238).
As an illustration of the iterative nature of mixed methods, an investigation employing mixed methods was conducted on the 1940 Ernest Hemingway novel For Whom the Bell Tolls. Specifically, qualitative GIS, incorporating non-quantitative data into a geographic information system, was used to analyze toponyms (place names, usually referring to specific topography) and to visualize patterns of how geographic locations in Spain were mentioned throughout the novel. The structure of the novel was divided into three rhetorical sites, or heteroglossia (multiple viewpoints expressed in a single literary or other creative work), consisting of the main plot (the “fabula”); the discursive plane, consisting of dialogs between the characters; and a monologue plane, revealing the thoughts of the Robert Jordan, the main protagonist. The iterations consisted of the initial qualitative step of close reading. In the subsequent quantitative step, data were collected on each mention of a location, which was subsequently assigned to one of the sites, and incorporated into the GIS, using a different GIS layer for each of the different sites. The next step consisted of geostatistical analysis to study relationship of the mapped geographic locations. Distinct differences in the patterns of locations in the three different rhetorical planes were found – locations were more geographically spread out in the monologue plane, and simultaneously more clustered than in the other two sites, which consisted of more concentrated clusters. These results prompted another close reading of the novel, with the conclusion that Jordan considered himself to be a “pan-Spaniard”, as suggested in the broad, large cluster in the monologue site (de Sá Pereira, 2019). This case study demonstrated one possible approach to employing mixed methods, using either qualitative or quantitative approaches at different stages of the analysis, depending on which method(s) is/are warranted at each stage.
In using mixed methods, quantitative, digital approaches are employed for the benefit of humanities scholarship, but Big Data and digital techniques do not become the focal points of scholarly output. They promote “digital hermeneutic (interpretative) strategies” and make room for the argumentation so vital in humanities research. In contrast to an exclusively technology- and algorithm-based focus on “distant reading”, such digital hermeneutic strategies facilitate “close reading” of a small number of texts, or even a single text (de Sá Pereira, 2019).
As digital humanities scholar Johanna Drucker states in an email conversation with art historian and scholar Claire Bishop: “When we ran the institutes, the single most interesting exercise was taking the participants from a research question into ways to address it through structured data and analysis. We could see pretty quickly that some projects were mapping projects, some were network analysis, some were repository building, etc. None of that is particularly interesting. The execution in most platforms is so formulaic it feels pointless to me—unthinking and uncustomizable. But the analysis of a complex problem into a data model – that is a really interesting intellectual exercise. Not the computational tools and visualizations per se, but the insights and potential new research questions that arise from them are” (Drucker & Bishop, 2019).