Signs of the Fleece

In Judges 6 of the Hebrew Bible, an angel of the Lord appeared to Gideon (who was, at the time, threshing wheat) to enlist the young man into service delivering Israel from Midianite bondage. Gideon proved to be a reluctant warrior. “How can I save Israel?” he asked; “my clan is weakest in Manasseh, and I am the least in my father’s house” (Judges 6:11). Gideon needed a tangible sign to know that it was truly the Lord who was speaking to him. And he got that confirmation when the angel lit his offering of meat and bread on fire. The miracle was proof enough for Gideon to start the subversive work of destroying the altars of Baal, though his people were not yet convinced.

The problem of belief is a persistent thread in the Pentateuch, whose characters must, on several occasions, distinguish between true miracles and mere trickery. So too the Pharaoh distrusted Moses and his brother Aaron, when Aaron cast down his rod that then turned into a serpent. Pharaoh’s own magicians were able to work similar miracles by apparent legerdemain (Exodus 7:8). Gideon’s fire could have been faked as well. On the evening of the battle with Midianites, Gideon then asked for another miracle, whether it was to solidify his faith or to steel his troops for upcoming battle. “I shall put a fleece of wool on the threshing floor,” he said. If, by morning, the dew collects on the fleece only and “it is dry on all the ground, then I shall know that You will save Israel by my hand, as You have said.” By the morning the dew collected on the fleece alone, just as he asked. One would think this would conclude the ordeal, yet Gideon timidly asks for another sign. “Let not thy anger burn against me,” he says. “Let me make this trial only this once with the fleece; pray, let it by dry only on the fleece, and on all the ground let there be dew” (Judges 6:33). So it was and Gideon’s men go on to scatter the Midianite armies.

Why weren’t Gideon and his people satisfied with the first miracle of the fleece? Once, in a conversation about game theory and market design, Alvin E. Roth, the Nobel laureate economist from Harvard Business School, suggested that this was the birth of the experimental method. For all Gideon knew, the dew always collects on the fleece. A point of comparison was needed to spot the difference between fraud and miracle.

The move to repeat an experiment in order to isolate the meaningful results from natural contingencies is at the heart of science, but also at the mimetic roots of comparative literature. To restate after Derrida and Saussure: meaning only makes sense in difference. The first skin was needed to establish the background against which the absence of the dew on the second skin could become visible.

Visible Absence

I came to comparative literature during my undergraduate years in the 1990s, as a refugee from the University of Michigan’s political science department. At the peak of high theory, the field seemed to pose a vibrant alternative to the arid analytics of philosophy and political science. There was something joyful and subversive in discovering the writings of Gayatri Spivak and Edward Said, in reading Cixous, Kristeva, and Zizek, in speaking in hushed tones, like conspirators, about Hegel and Husserl. This was a discipline for refugees, displaced in land and thought, theorizing displacement.

Years later, around the time of the last ACLA report, Terry Eagleton could meaningfully write about time “after theory,” in what Richard Rorty would call “the ebbing tide,” and Spivak, simply, “the death of the discipline” (Eagleton; Rorty; Spivak). The end made sense, for the displaced cannot long stay domesticated. To retain their identity they must (by definition) abandon their cozy mental abodes, even those of their own making. We could name this necessity to move “negation” along with Sartre, or we could call it “the poetics of estrangement” along with Svetlana Boym and Victor Shklovsky (Boym). That move sideways (again Boym), into the continuing retreat from “positivist” structures of knowledge formation defines the methodological dynamics of the discipline. The retreat is both cause and effect of a perpetual crisis affecting literary studies broadly, felt more acutely at its intellectual origins in comparative literature.

The trajectory of digital humanities should not be considered in isolation from these broader dynamics. The movement unfolds in several directions. Stylistically, it is a response to the extravagant excesses of high theory: its pomp, star culture, jargon, and ultimate alienation from the world. Semantically, the digital humanities expand on the tradition of structuralism, materialism, and pragmatism (in the sense of evaluating belief in terms of its practical application). Neither “digital” nor “humanities” does justice to the transformative program of the project. The dialectical pendulum swings from thinking to doing and from theory to method (and not at all from analog to digital, as the marquee would suggest). The digital humanities usher literary studies from the fertile age of speculative thought and model building to a time of instrumental reason and toward robust empiricism. It makes sense, then, to observe the concomitantly emerging academic interest in material contexts of all kind: in labs and maker spaces, in book history, in sociology of literature, in physical computing, in problems of regional inequities of access to information, in the future of books, presses, and libraries, in free culture, in the praxes of remix and remediation, and in the actual labor conditions at the base of our academic practice. Viewed in this light, the digital humanities extend rather than curtail the theoretical program of the previous generation.

Although few comparativists call themselves digital, the overall effect of the broader cultural shift is well reflected in our departments. DH is full of refugees too, and because it performs yet another conceptual move sideways, it can find a fitting home in the field of comparative literature. The inevitable backlash (as evidenced by several worried-sounding pieces in popular media) has already began (see for example Marche; Fish). Those who understand the history of the discipline—its intrinsic tensions between structuralism and post-structuralism, immanence and transcendence–will stay lithe and not make themselves too comfortable.

Knowledge Domains

What defines an intellectual field? In Gideon’s experiment, it is the fleece that makes observation possible. The ground appears undifferentiated otherwise. By unfurling the fleece on the ground, Gideon delimits a zone of topographic exclusion where purposeful activity can become distinct from arbitrary forces of the universe. The dew must collect here and not there–inside and not outside of the skin boundary. The second exclusion is temporal. Divine intervention is not necessary at all on that first night. Gideon merely needs to obtain a “read” – any read – against which subsequent observations can become meaningful. Michael Coogan, a prominent scholar of the Old Testament, reports that the name Gideon literally means “hacker” – which can be interpreted as a nickname for “the one who cuts down forbidden ritual objects” (Coogan). The story of Gideon, as told in the book of Judges, can be understood as Gideon’s transformation from a layman (“the least in my father’s house”) to a man of service, known for his military prowess, leadership, and iconoclasm. That change unfolds (literally) on a field where knowledge and belief are contested in differentiating “hacks” of time and space.

In the language of domain analysis, an approach to the study of information embedded in discursive communities, the colloquial “intellectual field” is better expressed as “knowledge domain,” understood broadly as an assemblage of people, ideas, institutions, and practices.[1] For the purpose of this paper, I represent the knowledge domain at hand more narrowly, as a list of citations extracted from a complete corpus of 8614 articles appearing in print in Comparative Literature between 2004 and 2014. Using Python and Sci2 I normalized the citations to unique author names and article titles in order to produce a table of relationships representing the citation network (Sci2 Team).

Formally, the network is comprised of “nodes” and “edges.” A node represents a single text: one either appearing in the corpus directly (published in Comparative Literature) or one merely referenced in the journal. Consider the case of a hypothetical article on Shakespeare’s Hamlet citing the play itself and other relevant scholarship (let’s say by Harold Bloom and Stephen Greenblatt)–scholarship that does not appear in Comparative Literature. In this example, an edge would be drawn between the original article, Hamlet, Greenblatt’s, and Bloom’s work. Since we include only the bibliographic data culled from Comparative Literature, sources “secondary” to the original article can connect to each other only through other articles native to the corpus. We cannot tell, for example, whether Greenblatt cites Bloom by looking at this network, since neither text was published in the journal. By this property, native sources appear as constellations of references. Frequently-cited sources appear larger and darker on the resulting map (Fig. 1).

 

Figure 1: Citation network detail, Comparative Literature 2004-14.

Despite its apparent formalization, network analysis far more resembles an art than a science. Initially, the network loads as an undifferentiated mass of nodes and edges (Fig. 2). In subsequent processing, I cluster the space using several layout algorithms. To derive Figure 1 from Figure 2 above I used Force Atlas 2, which is a physics-based layout, particularly suited to exploratory data analysis (Mathieu). Force Atlas treats nodes as particles that repel each other, where edges act as springs that bring their nodes closer together. The algorithm strives towards stasis in a stochastic way. No two visualizations of the same network will ever be the same. Moreover, Gephi, the software used to produce these layouts, exposes a number of interfaces through which I am able to “fine tune” the physical properties of that universe (Bastian, Heymann, and Jacomy). This has the potential to derive wildly differing visual results from the same dataset.

Formalization, analysis, and representation are therefore all involved in the interpretive process. The reader must be particularly careful in construing the natural metaphors used to describe the resulting model. For example, we could say that the physical distance between any two nodes in our graph represents some notion of “intellectual affinity.” Yet, distance between nodes in these diagrams is both non-deterministic and arbitrary. A different clustering algorithm could be used to bring any two given nodes closer together, if only in appearance. Much more work, quantitative and qualitative, would need to be done to formalize and to interpret our colloquial intuitions about “intellectual distance.”

Figure 2: Citation network before clustering. Comparative Literature 2004-14.

Clusters emerge as the system begins to stabilize during layout. The combined forces of attraction pull strongly connected, frequently cited papers towards the center of the map. Weakly connected, rarely cited papers float to the periphery (Fig. 3). In exploring the resulting maps, I assume, uncritically for now, that it is good to be cited. The discipline is, in many ways, defined by a shared archive of texts and references. Authors likely want to find themselves at the center of this graph. Conversely, life looks lonely at the edges. From these remarks, I intuit that connectivity in general is a common good. It is hard to imagine an intellectual field of isolates. For example, the sparsest network would consist of papers collected at random from disparate disciplines. What are these isolates then, towards the edge of the graph? They could be papers on esoteric themes, just entering the field. Alternatively, the isolates could represent papers written by guest authors from other fields, or just papers that, for some reason, do not share any common references with the rest of the network. To evaluate these conjectures we would need to contextualize isolates on a case-by-case basis, moving from distant to close reading levels of analysis.

Figure 3: Center-periphery relations within the network, Comparative Literature 2004-14.

What holds a field together? Another exploratory observation confirms that, in this domain, sources rarely connect to each other directly. Rather, source articles are held together by the connective tissue of what I would like to call “tertiary” nodes," which, in effect, comprise the common disciplinary canon. A tertiary node bridges two documents that would otherwise drift apart under the physical conditions of our model. Reducing the map to document titles only, while bringing the most connected nodes to the front, helps us isolate the results (Fig. 4). The reader will no doubt recognize Benjamin’s Arcades Project, Derrida’s “Politics of Friendship, David Damrosch’s What is World Literature?, Pascale Casanova’s World Republic of Letters, Franco Moretti’s Graphs, Maps, and Trees, and most prominently, Gayatri Spivak’s Death of a Discipline among the titles.

Figure 4: Network map, isolating the connective tissue, Comparative Literature 2004-14.

In Figure 5, drawn from papers appearing in the first ten years of the journal’s history (1975-85), the situation differs dramatically. The tertiary mesh of the time reflects a domain held together by primary materials: works like Aristotle’s Poetics, Homer’s Iliad, Milton’s Paradise Lost, Vergil’s Aeneid and Georgics, Augustine’s Confessions, and Ovid’s Metamorphosis. In the intervening 40 years of the journal’s existence, the shared literary canon underwent a substantive change from “great books” to theory and related meta-discourse reflecting on the state of the discipline. As a whole, the earlier network is also more weakly connected. In traversing the nodes we encounter more isolates–papers that do not share sources with any other documents in the corpus.

Figure 5: Tertiary nodes, Comparative Literature 1975-85.

Despite the increase in connectivity over time, my initial findings suggest that the domain of comparative literature is connected more weakly and perhaps in fundamentally different ways than other fields such as sociology, anthropology, history, and economics (Fig. 6). Unlike many of the journals examined in my forthcoming study on growth and decay of shared knowledge, Comparative Literature exhibits monolithic clustering around a single set of common texts, where in other disciplines we often see multiple well-defined clusters. What do these finding mean normatively? Should we strive for more connectivity and for more pronounced, diverse clusters? Should we rely less on a narrow canon of theoretical texts? It seems reasonable to suggest that connectivity is a common good, that researchers should read and quote each other and that they should, even in disagreement, share a set of theoretical or methodological assumptions. But these postulates are open to contention. The principles of domain analysis encourage us to place quantitative insight within its social, historical, and economic contexts. Citation networks form on the surface of a deeper culture. They are symptoms of the ways we teach, train, and communicate in our universities, departments, and professional societies.

Figure 6: Micro- and macro- economics in Econometrica 2001-11.

The way of comparative literature, like Gideon’s experimental hacking, has always been to advance through axes of contrast and correlation, by period and geography. The digital humanities offer yet another axis that bisects familiar concepts–in the case of this essay, ideas of canonization, center-periphery relationship, marginality, and reading–along a methodological divide. At its worst, the methodology devolves into shallow futurism and blunt instrumental reasoning. At its best, DH is a force of iconoclasm, used to question and to refine the prevailing orthodoxy.