My9s

“Elevating Influence”: Victorian Literary History by Graphs

Jana Smith Elford, Susan Brown, Michael Bauer, Jennifer Berberich and Jonathan Cable

Jana Smith Elford, University of Alberta, English; Susan Brown, University of Alberta and University of Guelph, English; Michael Bauer, Jennifer Berberich , University of Western Ontario, Computer Science; Jonathan Cable, University of Western Ontario, Computer Science@

Introduction

In the past four decades, largely due to the influence of feminist criticism, women’s literary history has made enormous advances in re-evaluating and recovering women’s writings in the Victorian period. The suppositions that women writers are limited to a separate sphere of influence or that they have not met the standard for admission into the literary annals have been fundamentally challenged.@ Yet, at the same time, traditional methods of conducting literary history have been severely contested. Histories, encyclopedias, and companions documenting a sustained, cohesive narrative have been regarded with an air of skepticism or contempt. Meanwhile, alternative, or “new” histories that attempt to undo the trend towards cohesive narrative have similarly been the object of critique: Harvard’s 2009 A New Literary History of America, for instance, despite the fact that it contains over 1,100 pages and two hundred chapters, was criticized because it chose “to fetishize particular moments” and thus risked “losing any coherent sense of how larger cultural forces operate” (Giles 627). Histories of the Victorian period have not been exempt from similar criticism: a recent reviewer for Victorian Studies declared that any “[c]ollection, even when focused on a limited topic and even when essays are written expressly for the volume, [is] notoriously prone to problems of coherence” (Anger 359). These contentions have strongly challenged traditional methods of doing literary history; indeed, how literary history might successfully be done has been the subject of debate now for decades (see Hume; Schellenberg; Perkins).
Despite claims by some that “literary history is pretty much [nonsense],” alternative means of doing literary history are emerging (Hume 632). With the hope of creating innovative ways of conducting literary scholarship, digital initiatives in literary history are offering a wealth of information, in terms of both the new discoveries consequent on the wider availability of primary texts previously available only through costly visits to archives and the development of new forms of born-digital scholarship that can lead to new methods of inquiry. Taken together, the explosion in the number of texts and the amount of contextual information available, combined with the emergence of new tools designed to help scholars process massive quantities of information, has the potential to shift dramatically the way we engage in literary history. However, these changes give rise to the challenge, among many others, of navigation: how to make vast amounts of primary or secondary information in digital form susceptible to scholarly inquiry is the increasingly pressing concern.
This paper considers the potential of network graphs as a method of investigating women’s literary history in the Victorian period and explores how visualization tools may be used to conduct what Franco Moretti calls “a more rational literary history” (Moretti 4). Building on Moretti’s contention that “deliberate reduction and abstraction [is] … a specific form of knowledge” (1), we present here some reflections on the use of a prototype tool for graphing literary history, OrlandoVision, which suggests alternate means of thinking of both the depth and breadth of the history of women’s writing.@ Developed to exploit the wealth of biographical, critical, and historical information about women’s writing in the British Isles embedded in the born-digital textbase of Orlando: Women’s Writing in the British Isles from Beginnings to the Present (Brown, Clements and Grundy), it has the advantage of being a digital tool whose abstractions are always dynamically linked back to the historical detail from which they are produced. This paper emerges from our exploration of the tool in conjunction with a facet of one team member’s current research, which explores some of the intersections between the Theosophical Society and feminists' political interventions at the Victorian fin de siècle. It therefore focuses on the visual representation of relationships and patterns of political activity within the field of late-nineteenth-century feminist writing. We have taken for our title a phrase used by Charles Hemans in praise of Jane Williams’s 1861 biocritical study The Literary Women of England, as a gesture both towards the dense interlinkages and webs of influence that existed within Victorian literary culture and towards the practice of elevating, or abstracting, relations of influence into prominence through the process of computer visualization.
Our approach involves drawing on data from Orlando in order to graph patterns of relationships among individuals involved with the Theosophical Society. Graphing in this context should not be confused with the graphing of points on a Cartesian plane (represented by x-y axes). Instead, the graphs here (referred to variously as directed or undirected graphs, or link-node diagrams) represent the connections between objects, called “nodes” or “vertices” and conventionally represented by dots, by means of “edges” usually represented by lines or curves.@ Although our tool is still very much in development, our results suggest that the potential for the application of digital graphing methods to literary history enables what Moretti calls “distant reading” to operate in conjunction with historical specificity and detail (1). Furthermore, while OrlandoVision responds to users’ own research priorities, it enables the exploration and discovery of relationships unforeseen by the scholar, which one might not have been able to search for using the methods already available through Orlando’s current online interface.
This paper will first give a brief background on Orlando and discuss some of the challenges facing the current first-generation digital methods of conducting literary history. Next, it explains some relevant features of the graphing tool, details our experiences with ‘reading’ OrlandoVision with a particular set of research interests in mind, and discusses the implications of these findings, providing screenshots and a short video of the OrlandoVision tool. We conclude with reflection on what is at stake in a literary history that relies on such tools in order to grapple with some of the problems produced by the digital turn.

Background: Orlando: Women’s Writing in the British Isles from the Beginnings to the Present

The Orlando Project began as an experiment in literary history, exploring the potential of computers to support new modes of research in the humanities, and in particular to exploit the possibilities of digital technologies for interpretive and critical scholarship. The resulting textbase constitutes the single most extensive and detailed resource in the area, hailed by the Modern Language Association’s Guide to Literary Research as “a model for similar databases that will supplant printed literary dictionaries, encyclopedias, and handbooks” (Harner 10). Though Orlando resembles a reference work, its electronic structure embeds a critical and theoretical framework to support advanced literary historical enquiry. Composed of eight million words of literary historical information in the form of more than 1200 critical biographies, mostly of women writers in the British Isles, as well as historical and bibliographical materials, it is extensively encoded using a tagset which categorizes and structures the prose with more than 250 tags for everything from cultural influences to authors’ narrative techniques. Because it is structured according to consistent principles, the Orlando encoding system enables a degree of cross-referencing and textual inter-relation impossible with print scholarship. However, the search-and-retrieval model of the current interface for Orlando, while user-friendly to the extent that it resembles first-generation online research tools, cannot leverage this encoding fully, particularly since search interfaces only find what the user asks for. Orlando seeks to exploit the power of second-generation text-exploration methods such as visual representation. As Moretti argues in the context of using visualization for inquiry into book history, the strength of such methods avoids “asking only those questions for which we already have an answer” (Moretti 26). Graphs permit us to see both interesting anomalies and general trends or patterns. Being digital, Orlando has the added advantage over print graphs of making it possible at any point to dive back into the source material to see the specifics from which the visual representation is produced.

‘Reading’ OrlandoVision:

In order to graphically represent the relationships embedded in the Orlando textbase, OrlandoVision uses a conventional network graph to represent relationships or associations, wherein the individuals are shown as nodes or vertices, and the relationships as edges or lines connecting the nodes. Users can visualize the entire Orlando textbase or a filtered subset which includes: literary period or date range, keywords, or, excluding all but writers with an entry in the Orlando database. [See Figure 1: Dataset Selection Frame] We filtered the data using the key word “theosoph,” which then pulls all entries in the Orlando database with any mention of theosophy, the Theosophical Society, or a variation thereof. This dataset was chosen for two reasons: first, because we wanted to test the extent to which the tool could be used to explore the general cultural, political, and religious influences on a particular group of women writers at the fin de siècle, and second, for practical reasons, because the small dataset enables the tool to process more quickly. After specifying a dataset using filters, the graphing tool then selects all entries and names connected to that particular request. These names are grouped, as nodes or dots, around the name of an entry to which they are linked, and joined with lines to this and all other names to which they are connected. [See Figure 2: Initial Graph of Theosophy Dataset] Initially, the lines of the Theosophy Graph are simply white, to reflect all connections in the subset of entries. This enables the user to see every connection between each author or individual. It is immediately evident which entries contain the most connections: larger starbursts and heavy, thick lines indicate the heavy-weights in this selection of data. In order to begin to make sense of this mass of lines, users can select a particular tag or group of tags present in the data, which provides semantic contexts or categories for the links it displays. [See Figure 3: Tag Selection Pane] Here we selected the tags we wished to bring to the fore in the visualization, and the colours for each of the tags. The tag selection made was based on several factors. First, the prevalence of nested tags within the Orlando data necessitated that we hide certain tags in order to avoid duplication of some connections. We made certain, then, not to select some of the more general tags which pull information that could come up in other tags, such as <reception>, <production>, and <textualfeatures> tags. Instead, we chose more specific subtags or children of these tags. The <response>, <fictionalization>, <penalties>, <recognitions>, and <sheinfluenced> tags, for instance, all represent a particular kind of critical literary reception.
Second, because some tags offer richer content than others in the dataset we selected, we deselected those that occurred infrequently. There are various reasons for this paucity of content: since the edges are all drawn between particular individuals’ names, and certain tags may not typically contain names, some tags are less represented in this graph than others, even though they may embed important connections of a different nature, such as links to texts or organizations. In addition, because OrlandoVision currently only makes connections through individuals, links between people by way of an <organizationname> , <genre> or <themeortopic> tag to which they are both linked are not represented in the graph. Although absence can be as revealing as presence, this winnowing out of the tags had the virtue of limiting the scope of the graph and focusing the visualization.
Once we had eliminated tags based on these criteria, we had a readable graph which brought surprising relations to the fore. [See Figure 4.0: Graph of Theosophy Dataset with Winnowed Tags] The graph displays several crossing lines which represent certain tags, and a prevalence of light blue, pink, and yellow, which represent respectively the tag <responses> (reactions to a literary work), <politics> (all information concerning a person’s political life), and <family> (all related to a person's family life, including family members, marriages, etc.). [See Figure 4.1: Graph with <responses>, <politics>, and <family> Tags Selected]. The predominance of the tag was not unexpected: it reinforces a strong trend that Joy Dixon describes in terms of “the connections between spirituality and politics in feminist political culture” (3). In the late nineteenth and early twentieth centuries, the Theosophical Society held considerable appeal for individuals with alternative political and religious beliefs who desired to practice an unorthodox spirituality. Those who had need for the reconstruction of a faith challenged by physics and biology, particularly Darwinism, were increasingly drawn to this syncretistic religion which emphasized the "universal brotherhood of humanity" and combined religious and scientific principles (Dixon 3). Feminists who were excluded from equal participation in traditional orthodox religion pursued membership in the Theosophical Society in high numbers, since it offered women an equal part in the enterprise "without distinction of … sex" (Dixon 3). At the turn of the century members of the Theosophical Society believed that they were about to lead the world into a new spiritual and political age (Dixon 8), and for many this object took a particularly feminist inflection. For these women, female emancipation was seen as “the first step towards the literal redemption of the nation,” (Dixon 179) and, eventually, the redemption of "all humanity" (Dixon 182). This objective meant that women involved with the Theosophical Society were politically engaged in a wide range of activities in order to effect social change.
To investigate whether the Theosophy graph reveals a trend towards feminist political action, we selected the <politics> tag on its own. The graph evidenced multiple, thick lines connected to two central nodes: Emmeline Pethick-Lawrence and Storm Jameson. [See Figure 5: Graph with <politics> Tags Selected] Upon closer inspection we discovered that most of Storm Jameson’s nodes ‘float,’ that is to say that they are not connected directly to other nodes in this subset of the Orlando data, so we highlighted@ Pethick-Lawrence in search of political interconnections within this group.[See Figure 6: Graph of Emmeline Pethick-Lawrence in Highlight Mode] This selection revealed that Pethick-Lawrence is directly connected to three writers, namely Charlotte Despard, Mary Gawthorpe, and Annie Besant, and one degree removed from four others, Henrietta Müller, Mona Caird, Eva Gore-Booth, and Storm Jameson. Yet the most interesting facet of this graph is the nearly self-contained web of interrelation that is rendered visible when the nodes one degree removed from each of the writers are selected. [See Figure 7: Graph of Pethick-Lawrence in Highlight Mode with Nodes Linked] Pethick-Lawrence, for instance, is connected directly to Emmeline Pankhurst: the text on which the link is based reveals that Pethick-Lawrence joined Pankhurst’s militant suffrage movement. Yet selecting Pankhurst in this dataset reveals only connections to individuals with entry nodes already toggled or made visible, that is, Caird, Despard, Gore-Booth, and Gawthorpe.@ [See Figure 8: Graph of Emmeline Pankhurst in Highlight Mode] Reading the Orlando text reveals that Caird shared a cab with Pankhurst to attend the Women’s Social and Political Union procession of 21 June 1908; Despard became one of the leaders of the suffrage movement with Pankhurst and Fawcett; Gore-Booth was associated with Pankhurst on the Lancashire and Cheshire Women Textile and Other Workers' Representation Committee, while Gawthorpe wrote about the split between Pankhurst’s militant suffragettes and the more moderate wing. [See Figure 9: Text Feature] Here, the expected connections between politics and spirituality emerge in a web of influence within which the feminist is inextricable from the spiritual, intimately connected to and influencing social activities, political actions, and literary endeavors.
Returning to the graph of Figure 4.1, while the abundance of political connections supports initial assumptions, the incidence of the <family> tag suggests deeper interrelations through networks that were unforeseen. This brings to mind the question: To what degree did familial relations influence political action, and, how does this relate to the social and cultural relationships which were not initially prominent in the graph? Selecting three additional tags provides one way of further mapping this terrain by means of comparison: <LeisureandSociety>, which relates to the cultural and social activities of a person; <IntimateRelations>, which details information related to emotional, psychological, material, or sexual intimacy; and <FriendsandAssociates>, a tag which references a writer’s friends and personal connections. [See Figure 10: Graph with <politics>, <leisureandsociety>, <friendsassociates>, <intimate relations>, and <family> Tags Selected] The visual comparison here reveals a complex and intersecting web of influence.
In addition to allowing discovery of both interesting anomalies and general trends or patterns, the OrlandoVision tool enables the user to trace these abstractions back to historical detail. Here, we chose to highlight the graph where the political and familial connect in order to pursue our question about the relation between political action and familial influence. [See Figure 11: Graph with <politics>, <leisureandsociety>, <friendsassociates>, <intimate relations>, and <family> Tags Selected in Highlight Mode] The node at which they intersect is W.B. Yeats, who is connected to Ezra Pound through <family>, who is in turn connected to H.D. through both <politics> and <intimaterelations>. These links are based on the fact that Pound was Yeats’ best man at his wedding; the text of the database, accessed through the Text feature of the graphing tool, reveals the detail that Pound also had a lasting romantic relationship with H.D. that broke off at least in part due to his anti-semitism and support for fascism (H.D. herself was an anti-fascist). This pair of connections also demonstrates a couple of specific features of OrlandoVision. First, it is important to understand the way in which the tagging works within a particular set of materials: as this example makes clear, the <family> tag encompasses not only familial relations but in fact accounts of family events and can include the names of non-family members. This makes it imperative that the tool permit the researcher to access the textual content of the links. Second, this example reveals that two very different tags may contain information about one relationship that is multifaceted and complex. In this case, the relationship between Yeats and H.D. is characterized within Orlando by at least two dominant features: their intimate relationship and their political disagreements. Furthermore, their mutual connection to Theosophy suggests additional spiritual interrelations which beg to be explored. This particular graph of literary history thus reveals a broader pattern of political-religious relations, a web of interlinkage which asks new questions and illuminates new avenues of exploration.
Although we had expected to find connections between feminist authors, political action, and theosophy at the turn of the century, we instead uncovered modernist writers (two of them male), friendships, and intimate relations nearly a generation later. While both theosophy and politics are still strong factors in this particular facet of the data, rather than definitive answers, a spate of questions arise: To what degree did the theosophical leanings of Yeats and H.D. influence their relationship to Pound? Their writings? What was Pound’s relation to Theosophy? Was the version of Theosophy espoused by Yeats and HD significantly different than the earlier version espoused by late Victorian feminists? How political was this Theosophy which H.D. and Yeats espoused? What was its relation to anti-fascism? Had Theosophy at this point become less about political action and more about personal spiritual revelation? These expansive questions, suggest, in Moretti’s words, a “widen[ing of] the domain of the literary historian,” insofar as the information we have found is more far-ranging than we initially imagined (2). Here, the OrlandoVision tool enables a broadening inquiry into women’s literary history even while focusing on a narrow subset of the graph.

Conclusion and Next Steps

Our findings using OrlandoVision reveal a complex interrelation of spiritual, political, and social influences on literary history in the Victorian period which offer the potential to challenge and expand the ways we conceived of feminism, spirituality, and how social and political movements function. While we had anticipated some patterns in the Theosophy dataset, the network of connections made evident in the graph possesses a depth which complicates and nuances our understanding of the nature of the interrelationships amongst these writers, as well as a breadth which stretches beyond literary periods. In this sense, the OrlandoVision tool expands how we engage with literary history and, in so doing, widens the domain of the literary historian, particularly in that it avoids “asking only those questions for which we already have an answer” (Moretti 26). The dynamic nature of the tool also allows the user to move from the abstract visualization back to the specifics of the text, thus enabling the operation of “distant reading” (Moretti 1) in conjunction with the details that have generated that reading.
What is stake in how we conceive of this kind of venture into a new mode of literary historical inquiry is the often unconscious ideological freight that can accompany digital tools, whereby the computer is viewed in its empiricist tradition, as the “means by which critical interpretations may be verified,” generated, or proven (Ramsay 167). It is easy to approach OrlandoVision as a tool which would enable the user to come up with “correct” interpretations, or to search for specific information, ignoring or side-stepping unresolved questions or anomalies within the data which may challenge assumptions. Yet, as we hope this account of using OrlandoVision in the context of a particular research inquiry demonstrates, the OrlandoVision representation of literary history, the graph drawn by the encoded text of the database, is itself still a text, and it is moreover reading a text that is the product of many and diverse contributors. This is to say that it is still located within what Stephen Ramsay calls “the rich tradition of interpretive endeavours” (167). Viewed as such, our activities with the tool become themselves a hermeneutic activity, an act of interpretation. Further work needs to be done to determine whether this interpretive activity involves the same pitfalls, follies, and opportunities to construct meaning as engagement with print texts, or new and different ones.
Jana Smith Elford is a 2010 Canada Graduate Scholar in the Department of English at the University of Alberta. Her current research examines the rhetorical strategies at work in the cross-genre writings of feminists at the Victorian fin de siècle. She has published her research in Orlando: Women's Writing in the British Isles from the Beginnings to the Present and is currently involved with user testing for the Orlando initiative, Text Mining and Visualization for Digital Literary History.

Works Cited

Anger, Suzy. “Rereading Victorian Fiction.” Victorian Studies 44.2 (Winter 2002): 359.
Barabási, Albert-László. Linked: The New Science of Networks. Cambridge, MA: Perseus Publishing, 2002.
Brown, Susan, Patricia Clements, and Isobel Grundy, ed. Orlando: Women’s Writing in the British Isles from Beginnings to the Present. Cambridge: Cambridge UP, 2006-2010. Online.
Christakis, Nicholas A. and James H. Fowler’s Connected: The Surprising Power of Social Networks and How They Shape Our Lives. New York: Little, Brown, 2009.
Dixon, Joy. Divine Feminine: Theosophy and Feminism in England. Baltimore Johns Hopkins UP, 2001.
Gilbert, Sandra M. and Susan Gubar. The Madwoman in the Attic. New Haven: Yale UP, 1979.
Giles, Paul. “A New Literary History of America.” Journal of American Studies 44.3 (2010): 627-628.
Harner, James L. Literary Research Guide: An Annotated Listing of Reference Sources in English Literary Studies, 5th ed. New York: MLA, 2008.
Hume, Robert D. “Construction and Legitimation in Literary History.” The Review of English Studies 56.226 (2005): 632-661.
Moers, Ellen. Literary Women. New York: Oxford UP, 1985.
Moretti, Franco. Graphs, Maps, Trees: Abstract Models for a Literary History. London: Verso, 2005.
Perkins, David. Is Literary History Possible? Baltimore: Johns Hopkins UP, 1992.
Schellenberg, Betty A. “Writing Eighteenth-Century Women’s Literary History, 1986-2006.” Literature Compass 4.6 (2007): 1538-1560.
Showalter, Elaine. A Literature of Their Own. Princeton: Princeton UP, 1977.
Ramsay, Stephen. “Toward an Algorithmic Criticism.” Literary and Linguistic Computing 18.2 (1 June 2003): 167-174.

“Elevating Influence”: Victorian Literary History by Graphs

Jana Smith Elford, Susan Brown, Michael Bauer, Jennifer Berberich and Jonathan Cable

Endnotes

1  Jana Smith Elford provided the use case for this paper and tested the prototype in connection with her doctoral research; Susan Brown guided the prototype design from a literary historical perspective; Michael Bauer directed the technical work; and Jennifer Berberich and Jonathan Cable did the coding of the prototype.

2  Early studies by Ellen Moers, Elaine Showalter, and Sandra Gilbert and Susan Gubar all paid significant attention to the position and works of nineteenth-century women writers. They have been followed by hundreds of studies of individual writers, considerations of movements such as the New Woman, reprints, scholarly editions, and anthologies of primary texts. Together these have remapped the Victorian literary landscape.

3 For a video of OrlandoVision in action, click here or visit: [http://vimeo.com/19941639 ]. Work on OrlandoVision has been generously supported by the Sharcnet High Performance Computing Consortium, which provided programming resources, by the Social Sciences and Humanities Research Council of Canada, and by the University of Alberta’s Faculty of Arts.

4  Two accessible studies of graph theory and its application to the analysis of social groups are Albert-László Barabási’s Linked: The New Science of Networks and Nicholas A. Christakis and James H. Fowler’s Connected: The Surprising Power of Social Networks and How They Shape Our Lives.

5 The examples given thus far have used theOrlandoVisiondefault Graph Mode, which shows the full graph. The remainder use Highlight or Toggle mode. Toggle and Highlight modes allow the user to focus further. Toggle Mode allows users to start with the lines and other names associated with a single node. When a user left-clicks on a related node, all the lines and nodes connected to that second node will appear. One can thus expand the visible portions of the graph incrementally, gradually exposing a subset of the full graph. Highlight Mode is similar to Toggle mode; the other nodes in the graph are visible but dimmed so that the toggled nodes are more prominent, allowing users to see the connections in the entire graph while still being able to focus on the connections that matter most to them.

6 This result suggests that those already highlighted comprise those theosophists who were active in this branch of the suffrage movement, although further inquiry would be necessary to ascertain whether there are theosophists involve with suffrage but not linked to Pankhurst. In addition, it must be remembered that because the dataset we are using here is limited to writers connected to theosophy, Pankhurst’s relationships to people absent from this subset cannot be represented in this graph; a graph of all her relationships withinOrlandowould be more extensive.

Links

Page 1

"here" http://vimeo.com/19941639

"http://vimeo.com/19941639" http://vimeo.com/19941639

"short video" http://vimeo.com/19941639

"Dataset Selection Frame" http://scdp.uky.edu/rosenman/0001.html

"Initial Graph of Theosophy Dataset" http://scdp.uky.edu/rosenman/0002.html

"Tag Selection Pane" http://scdp.uky.edu/rosenman/0003.html

"Graph of Theosophy Dataset with Winnowed Tags" http://scdp.uky.edu/rosenman/0004.html

"Graph with &lt;responses&gt;, &lt;politics&gt;, and &lt;family&gt; Tags Selected" http://scdp.uky.edu/rosenman/0005.html

"Graph with &lt;politics&gt; Tags Selected" http://scdp.uky.edu/rosenman/0006.html

"Graph of Emmeline Pethick-Lawrence in Highlight Mode" http://scdp.uky.edu/rosenman/0007.html

"Graph of Pethick-Lawrence in Highlight Mode with Nodes Linked" http://scdp.uky.edu/rosenman/0008.html

"Graph of Emmeline Pankhurst in Highlight Mode" http://scdp.uky.edu/rosenman/0009.html

"Text Feature" http://scdp.uky.edu/rosenman/0010.html

"Figure 4.1" http://scdp.uky.edu/rosenman/0005.html

"Graph with &lt;politics&gt;, &lt;leisureandsociety&gt;, &lt;friendsassociates&gt;, &lt;intimate relations&gt;, and &lt;family&gt; Tags Selected" http://scdp.uky.edu/rosenman/0011.html

"Graph with &lt;politics&gt;, &lt;leisureandsociety&gt;, &lt;friendsassociates&gt;, &lt;intimate relations&gt;, and &lt;family&gt; Tags Selected in Highlight Mode" http://scdp.uky.edu/rosenman/0012.html