Network Theory, Plot Analysis
QUANTITY, FORM. How does a literary historian become interested in empirical study. For me, it began with evolution, and specifically with Ernst Mayr's theory of speciation, where geography plays a large role in the origin of species. This made me try to make literary maps, to study the origin of genres, and maps turned out to be much better when they were based, not on a single text, but on a series. So I began extracting data from novels, and this was my first modest attempt at quantitative work. Modest, because it was instrumental, the data were simply a means to the maps, and because the quantities were small: beginnings and endings in Austen: 12 data. Balzac's young men in Paris and their lovers: 20. Foreign concepts in Russian novels of ideas: another 20. Not much.
Book history was better: print runs, library holdings, translation flows were all much larger corpora. But, the problem was – they remained external to literary form. You could count how many Gothic novels were published in the 1790s, what libraries held them, where they were translated – none of this said anything about the Gothic as form. And for me that was the whole point; that's why I had become interested in evolution to begin with: it was such a powerful theory of how forms change. And instead, when I found interesting formal patterns, as in some maps, quantities were small; when quantities were good, form disappeared. Frustrating.
This, in the 1990s. Then, a few years ago, I began to apply the quantitative approach to – style. Initially, in a very elementary way: frequency searches; at times even those are lucky, in Jude the Obscure, it turns out, the most frequent word is "Jude" [with 814 occurrences], then comes "said", [783], then "Sue" [511], and the fourth most frequent word is "don't" [418; in Middlemarch, to give you an idea, "don't" is the thirtieth most frequent word]. And it's striking, the repressiveness of a whole novel compressed into a word but, first, such luck is rare, so one can't really count on it, and, second, you don't do quantitative work in order to replace subjectively chosen details with statistically significant ones – you do it in order to go beyond individual details.
For me this happened recently, with a study of 7,000 titles of British novels between 1740 and 1850, where the first and fundamental finding was indeed quantitative, i.e. how much shorter titles suddenly became in the late eighteenth century: from between 10 and 20 words per title in mid-century, to 10 around 1770, and then 6 by 1790. Within this general trend, the most radical changes occurred at the two ends of the spectrum: long titles with 15 or more words collapsed from over 50 to 10% or less, while very short titles – with only 1, 2, or 3 words – multiplied from 5% in the 1740s and 50s, to around 25%. And one type of short title – abstractions, like Persuasion, or Pride and Prejudice – was particularly important: titles used to tell readers something about the plot of the novel – a summary of events, the name of the heroine, the setting – but Pride and Prejudice did none of this, it highlighted the point of the novel, and this matters, because it changed the horizon of expectations around the form. That titles became shorter is interesting, but, in the end, So what. That by being shorter they found a rhetorical strategy that made readers look for a unity in the story – this was a perceptual shift. In which – finally – quantity and form clicked: a title with 2 words was not the same as a title with 20, divided by ten: it was different. It worked differently. When quantity changed, so did form.
A NEW OBJECT OF STUDY. Applying a quantitative approach to style not in order to replace one detail with another, but to turn details into series; and then, take these series as a new object of knowledge. Eventually, you end up with something I will now explain. Let me explain what this is, first of all. At Stanford there is a linguist, Chris Manning, who has developed a program that reads a corpus and "tags" each word specifying what part of speech it is – determiner, preposition, adjective, noun, etc. – and there is a small group led by my colleague Matt Jockers who have formatted about 300 19th-century British novels in the way the tagger needs. We fed the novels to the tagger, and got them back subdivided into grammatical categories: IN, subordinating prepositions and conjunctions [astride, among, whether, out, despite ...] constituted 11.7 percent of all words in the 1780s; 12.5 in the 90s; 12.8 in the 1800s; NN, nouns, were at 12.7, 14.1, 14.1; and then NNP, proper nouns, RB, adverbs, VBD, past tense of verbs, VBZ, third person of the present – you can follow the ups and downs of all these categories across time.
A new object of knowledge ... which has nothing in common with texts anymore. Texts obviously remain the "real objects" of literary history, that's what we read, yesterday on the plane I was reading Hedda Gabler, not this. But real objects are not the same as objects of knowledge, just as reading is not the same as knowing. We want to read, Ibsen; we want to know, something like this.
But, what can one do with "this". I'll give you an example. I am writing a book on the figure of the bourgeois, there is a Victorian chapter, where I discuss the Victorian nude and claim that its real function was not that of revealing the body – and by implication material reality – but rather of veiling it; then I began to wonder whether language could also do a similar "cover up" – and I thought of adjectives. That’s what adjectives are for, in the end: padding, euphemism, camouflage... The tagger is perfect for this kind of questions, so I charted the results, ready for a rise of the adjective in mid-century, and – a perfectly flat 6% through the whole century.
Such a neat idea ... But quantitative data are great at falsifying hypotheses. Experiments are "a dialogue between fact and fancy", Peter Medawar has written: we are all good at fancy, in literary studies; quantitative data, are good at facts.
FROM STYLE TO PLOT. Word frequency, abstractions, parts of speech, adjectives ... [and in related research projects, function words, factor analysis, paragraph structure ...] Style: something that nowadays can be studied quite well in a quantitative fashion. Plot, is more difficult. And if you work on novels or plays that's a problem, because plot is the key: you can't quantify that, the whole project remains shaky. But – how can one do that.
Perhaps, via network theory. This is a theory that studies connections within large groups of objects: the objects can be almost anything – banks, neurons, film actors, research papers, friends... – and are usually called vertices; their connections are called edges, and the analysis has revealed many unexpected features of large systems – the most famous one being the so-called "small-world" property, or "six degrees of separation": the rapidity with which one can reach any vertex in the network from any other vertex. The theory proper requires a level of mathematical knowledge which I unfortunately lack; and it typically uses huge quantities of data which will also be missing from my paper. But even in my pre-modern version it has interesting things to say about plot; plus, this paper will soon be followed by a second one, co-authored with two research scientists at the IBM Visual Communication Lab, Fernanda Viegas and Martin Wattenberg, and with Matt Jockers at Stanford; and this second paper will have digital data collection, math, professional visualization – everything. But not today.
CHARACTER-NETWORK. A network is made of vertices and edges; a plot, of characters and their interactions: characters will be the vertices of the network, interactions the edges, and here is what the network from Hamlet looks like (fig. 5, to follow). There are questionable decisions here, mostly about The Murder of Gonzago, which we can discuss later, but, basically, in this figure two characters are linked if some words have passed between them: an interaction, is a speech act. This is not the only way to do things, the authors of a previous paper on Shakespeare had linked characters if they had speaking parts during the same scene, even if they did not speak to each other: so for instance, for them Osric and the Queen would be linked [because they both have speaking parts, and are on stage together], whereas here they are not, because they don't speak to each other. My network emphasizes explicit connections, theirs implicit ones, and is obviously denser, because they have all of my connections, plus some; both are plausible, and both have at least two significant flaws. First, the edges are not "weighted": when Claudius tells Horatio in the graveyard scene, "I pray thee, good Horatio, wait upon him", these eight words, which don't even receive an answer, have in this figure exactly the same value as the thousands of words exchanged between Hamlet and Horatio. This clearly can't be right. And, second, the edges have no "direction": when Horatio addresses the Ghost in the opening scene, his words place an edge between them, but of course that the Ghost wouldn't reply and would only speak to Hamlet is important, and should be made visible. But, I just couldn't find a non-clumsy way to visualize weight and direction; it's one of the many issues for the follow-up study.
Anyway. Four hours of action, that become this. Time turned into space: a character-system emerging from many character-spaces, to use Alex Woloch's concepts in The One vs. the Many. Hamlet's space (fig. 6, to follow): in red, all the direct links between him and other characters; Hamlet and Claudius (fig. 7, to follow): in red or blue, characters who speak to one but not to the other. Ophelia and Gertrude (fig. 8, id.): the small space of the only two women in the play. And so on.
But, before analysing some of these spaces – why use space to represent plot to begin with? What do we gain? For me, two things: every part of the plot is explicitly linked to all the others – and nothing ever disappears. What is done, is never undone. Whether a character is present on stage or not at any given moment, it is never not there in the network. Once the Ghost shows up at Elsinore, Hamlet's life changes forever, whether the Ghost continues to manifest itself or not. Once the network splits between Claudius and Hamlet (fig. 9, id.), characters who are linked to both [except for Osric and Horatio, whose link to Claudius is however very weak] are killed, whether Hamlet and Claudius are actually there or not: Polonius and Laertes are killed by Hamlet of course, and Gertrude by Claudius, but Ophelia, Rosencrantz, and Guildenstern, die when neither Hamlet nor Claudius are present. Killed by their position in the network, as it were.
MODELS, EXPERIMENTS. Now, a network of Hamlet is clearly not Hamlet: it's a model of Hamlet, and like all models, it's the result of a process of reduction and abstraction that makes it quite obviously less than the original object. But the model is also more than the original Hamlet, in that it makes us see some of the play's structures so much more clearly: this correlation of death with a specific region of the network, for instance: outside of the double link with Hamlet and Claudius, no one dies. Or, take the question of the protagonist. An application of network theory to the Marvel Universe has re-defined it as "the character that minimizes the sum of the distances to all other vertices" : one measures the distances among characters, makes the average, and finds who is closest to the network as a whole. The center of the network. In their case, it was Captain America: at an average distance of 1.70 from all other characters. In our case, Hamlet, at a distance of 1.45. One degree of separation between him and these 16 characters (fig. 10, id.); two degrees with the others (fig. 11, id.).
A model makes us see structures ... and it's not just a matter of seeing. One can intervene on a model; change some variables; make experiments. Think of the protagonist again. For us literary critics, the protagonist is important because it's a very meaning-ful part of the text; we would never dream of discussing a novel, and skipping the protagonist. But in a sense, this is exactly what network theory would do: take the Hamlet-network (fig.12, id.), and remove Hamlet, to see what happens (fig. 13, id.). And what happens is that the network almost splits in half: between the court and the part of the play that includes the Ghost and Fortinbras all that's left are the three edges linking Horatio to Claudius, Gertrude, and Osric: 36 words. If we used the first Quarto, the fissure would be even more visible (fig. 14-15, id.).
Why is the protagonist significant for network theory? Not in itself, not for what is "in" it, but as a way to test the resilience of the network. And resilience varies from case to case. Take Othello, and remove Othello (16-17); Lear, and remove Lear (18-19); Macbeth, and remove Macbeth (20-21). All of the networks are weakened, of course: but not in the same way as Hamlet. Macbeth is the clearest anti-type: removing Macbeth has a strong effect on a whole group of characters that become isolated, or almost isolated, from the rest of the plot; but not much on the general structure of the network. Macbeth has his own "periphery", down at the bottom of the network – but he doesn't act as a "hub", in the way Hamlet does. More than Hamlet, his equivalent would be the second most central character of the play: Claudius. In purely quantitative terms Claudius is actually almost exactly as central as Hamlet [average distance of 1.52, versus 1.45]; but not so in structural terms, when we remove him from the network (fig. 22-23), all that happens is a lesser version of the Macbeth case: individual characters are affected, at the top right, but the general structure not much. Even if we remove, first Hamlet (fig. 24), and then Claudius (fig. 25), the second diagram adds little to the first [or, more precisely, subtracts little from it]. But if we remove Hamlet and the third most central character, which is Horatio (fig. 26), not only do we get radical fragmentation, but the Ghost and Fortinbras – which is to say, the beginning and the ending of the play – are completely severed from the plot. To do the same to the structure of the other tragedies, we would have to eliminate seven or eight characters.
This is why models matter: they allow us to make experiments, and ask new questions. Why is the Hamlet-network so much more dependent on the protagonist than the other tragedies? Why is the removal of Claudius so insignificant? Why is Horatio so important?
RESILIENCE, CLUSTERING. Why is Hamlet the play so dependent on Hamlet the prince. Possibly, because of how sovereignty appears in the play. All of the tragedies begin with a figure of legitimacy, followed by an usurpation of sorts, and then by a second figure of legitimacy; but in Macbeth and Lear legitimate sovereigns are well connected to the rest of the network: Duncan and Malcolm [in red and purple] (fig. 27), or Lear, Albany and Edgar (fig. 28) have powerful antagonists, but the fields of legitimacy and usurpation are basically balanced. In Hamlet, no (fig. 29): between old Hamlet and Fortinbras on one side, and Claudius on the other, there is a total disproportion – and that's probably why the network must rely on Hamlet so much: the usual balance of power is not there, and Hamlet must fill the void. Why the balance is not there is a different question – on which network theory has probably nothing to say. That it is not there, is one of those things that models make visible.
Considering how overwhelming Claudius's presence is in this diagram, at first sight it's odd that his removal should affect it so little. The explanation lies in another concept of network theory, i.e. "clustering". "If vertex A is connected to vertex B and vertex B to vertex C", writes Mark Newman, "then there is a heightened probability that vertex A will also be connected to vertex C. In the language of social networks, the friend of your friend is likely also to be your friend." This is what clustering means: A and C connect, the triangle "closes", and the resilience of that part of the network increases (fig. 30), look at the space between Hamlet, Claudius, Gertrude, Polonius, Ophelia and Laertes: everybody is connected to everybody else: clustering is 100% – and that's why removing Claudius has such little effect: his main field of action is one where plenty of connections already exist, and so, whether he's there or not, things don't change that much.
Horatio is the opposite: he is significantly less central than Claudius in terms of distance from all other characters [1.83 as opposed to 1.52], but he inhabits a part of the network where clustering is low (fig. 31) and which is therefore more likely to disintegrate once he is removed. In this, Horatio is the perfect gateway to the periphery of the play: those least connected characters (fig. 32), with only one link to the rest of the network; at times, just one sentence. Very little. But as a group, these peripheral characters do something important: they point to the world beyond Elsinore: the gentleman, sailor and ambassadors who speak to Horatio, and one of the messengers to Claudius, are links to the "English" subplot; Cornelius and Voltemand, to "Norway"; Reynaldo, to Laertes's "France"; the Priest and Grave-digger, to the world of the dead. Centrifugal threads – "tendrils", as network theory sometimes calls them – which are not there in the other tragedies, and contribute to the uncanny sensation that Elsinore is just the tip of the tragic iceberg: geography playing a somewhat similar role to genealogy in Greek tragedy: the hidden dimension of fate. Genealogy rooted vertically, in myth; geography, horizontally, in something like the nascent European state system.
HORATIO. I may be exaggerating here, projecting onto this diagram Napoleon's words to Goethe at Erfurt – politics as the fate of modern man. But Horatio's space fig. 33 – ambassadors, messengers, the opening scene among the sentinels, and of course the final scene – this space seems to announce what will be soon called, not Court any longer, but State. The court, the space of 100% clustering we saw a minute ago – where one is always seeing others and being seen, as in Elias – that, in Hamlet, is fundamentally two families: Ophelia Laertes Polonius; Claudius Gertrude Hamlet. Horatio's world is much more impersonal: he has no links at all with Polonius, Ophelia and Laertes, and only a sentence with Claudius and Gertrude. I may be making too much of this; or, Horatio may really be a fantastic half-intuition on Shakespeare's part; and I say "half", because there is something enigmatically undeveloped about him. Think of Posa, in Schiller's Don Carlos. Don Carlos is to a large extent a remake of Hamlet, and Posa is certainly a remake of Horatio's: another lonely friend of another sad prince in another oedipal play. But Posa has a very good reason to be there: he is that new figure, so important for modern drama: the ideologue. There is something he wants to do. Horatio? Why is he there? Kent is near Lear out of loyalty; Macduff, near Malcolm to avenge his family. Horatio?
Horatio has a function in the play, but not a motivation. No aim, no emotions – no language, really, worthy of Hamlet. I can think of no other character that is so central to a Shakespeare play, and so flat in its style. But the style of the State will also be flat. And, except for the grave-diggers, this is also the kind of language we encounter at the periphery of Hamlet: the typical utterances of this region are orders, and news – "And we here dispatch/ You, good Cornelius, and you, Voltemand" [I.2.33-4]; "Sea-faring men, sir. They say they have letters for you" [IV.6.2-3] – and orders and news aim at reducing ambiguity, thus lowering the play's "figurality rate" [to use a concept of Francesco Orlando's]. Conversely, as we move towards the center of the network figurality rises: from Hamlet's hyper-compressed first lines to Claudius, to the soliloquies that occupy, so to speak, the center of the center. You see the possibility here: the structure of the network as a way to analyze plot and style: different uses of language emerging in different regions. We'll test that in the follow-up study.
SYMMETRY. Networks are made of vertices and edges; plot networks, of characters and verbal exchanges. In plays this works well, because words are deeds, and deeds are almost always words. In novels it's different, much of what characters do and say and even think is not spoken but reported by the narrator, and as a consequence direct discourse covers only a part of the plot – at times, a very small part. This makes the transformation of plot into network a lot less accurate; but it's too tempting to just let go, and so I will show a few networks from The Story of the Stone and Our Mutual Friend just the same. A couple of years ago I conjectured that the number of characters could be a major source of morphological differences between Chinese and Western novels, and networks seem to be a good way to test the idea.
Unlike Hamlet however, I won't present networks for the entire text, but only chapter-networks drawn from the first twenty chapters of each book, as I could perhaps manage Our Mutual Friend [even though, by Western standards, it has a lot of characters], but certainly not the hundreds and hundreds of characters of The Story of the Stone. In this novel, chapters have between 5 and 28 speaking characters each, with an average of 14.7 characters per chapter; in Our Mutual Friend, a lot less: between 3 and 14 speaking characters, with an average of 6.9. And here is a typical Dickens chapter: the first of the second book (fig. 34), which introduces Jenny Wren and Headstone; the following one, a variation on this (fig. 35), with Lizzie's other suitor, Wrayburn, and Jenny's father; the fourth, with the revenge of the Lammles over Podsnap via his daughter (fig. 36). And so on.
Now, in Western novelistic poetics, aside from a few neo-classical moments, symmetry has never been an important category. But you look at these Dickens networks – (fig. 37-43) – and it's stunning how symmetrical they are. Probably, there are two reasons for this. The first is that the building blocks are themselves often symmetrical: husband and wife, parent and child, brother and sister, suitor and beloved, friend and friend, rival and rival, colleague and colleague, employer and employee ... And then, second, these pairs can project their symmetry onto the chapter as a whole because there is very little "noise" around them – very few characters who can disrupt the symmetry. Or in other words: with few characters, symmetry emerges almost inevitably, even in the absence of a culture of symmetry.
A culture of symmetry is on the other hand extremely active around Chinese novels, with their "aesthetic expectation", as Andrew Plaks puts it, that
"the overall sequence of chapters [...] will add up to a round and symmetrical number, typically 100 or 120. The pronounced sense of symmetry [...] provides the ground for a variety of exercises in structural patterning. Most noticeable among these is the practice of contriving to divide an overall narrative sequence precisely at its arithmetic midpoint, yielding two great hemispheric structural movements".
Hemispheric movements: something like that exists also at the scale of the chapter, where it’s in fact announced by the couplets that function as epigraphs: "Zhou Rui's wife delivers palace flowers and finds Jia Lian pursuing night sports by day / Jia Bao-yu visits the Ning-guo mansion and has an agreeable colloquy with Qin-shi's brother". Character A does this and meets character B; C does that and meets D. The chapter is presented as divided into two halves, of equal weight. "A very earnest young woman offers counsel by night / And a very endearing one is found to be a source of fragrance by day". Perfect symmetry: "parallel prose", as Chinese aesthetics calls it. So you take the chapter-networks from The Story of the Stone, where black edges refer to the first half of the chapter, and red ones to the second half ... (fig. 44-50).
For me, this was fascinating, because I had assumed that Chinese novels would have far more symmetry that European ones. But, no. And the key is probably again the number of characters: just as with few of them symmetry emerges almost by itself, with many characters it becomes just too unlikely, at least at the scale of the chapter. But then, what do these patterns mean? Dickens's symmetries signify something quite clear: a melodramatic substratum of love or hatred, always ready to erupt. A-symmetry?
GUANXI. Let's look at one network in detail, the first half of the seventh chapter of the novel (fig. 51). Zhou Rui's wife, who is a member of the staff of the Rong mansion, must report to Lady Wang on the visit of a distant relative; she doesn't find her in her apartment, asks about her, is sent to other parts of the compound, is given some errands, inquires about some new faces as well as about people she hasn't seen in a while, is asked to intercede for her son-in-law ... and in doing all this she ends up meeting a dozen characters – or more exactly, speaking with a dozen characters, she meets at least twice as many, while another twenty or so are mentioned in the various conversations.
Now, nothing major happens here: people talk, walk around, play go, gossip... No single interaction is significant in itself. But if they are taken together they become significant, because they embody an informal reconnaissance and re-activation of relationships: making sure that the nodes in this region of the network are still communicating well: because, with hundreds of characters, the dis-integration of the system is always a possibility. In this chapter, the reconnaissance is motivated by Zhou Rui's wife being an employee of the mansion; elsewhere, the narrative reason is different, and brings us close to one of the most distinctive keywords of Chinese culture: guanxi: something like "connections", translate Gold, Guthrie and Wank; part of "a specifically Chinese idiom of social networks [...] linked to other building blocks of sociality such as ganquing [sentiment], renqing [human feelings], mianzi [face], and bao [reciprocity]": all in all, a world which is "neither individual- nor society-based, but relation-based". And one must work for these relations; they are not a given, they are an artifact; "manufacturing obligation", "chain of transactions", "indebtedness", "consciously producing" connections – this is the lexicon of guanxi.
A chain of transactions that generate indebtedness (fig. 52) in chapter 24 of the novel, Jia Yun, who is a poor relative of the Rong-guo house, is looking for work; he asks Jia Lian, but receives only vague promises; so he turns to his uncle Bu Shi-ren, who owns a store, hoping to get some perfumes on credit to use as presents. Bu Shi-ren says no, but invites Jia Yun to dinner – but then his wife makes a scene, and the invitation is dropped. Jia Yun then bumps into a drunk, who turns out to be his neighbor Ni Er, a racketteer, who lends him the money to finally buy the present and offer it to Xi-feng, who is in charge of the finances of the clan ... This is how guanxi works. A character rallies all its resources and allies in order to "manufacture obligation"; and this, inevitably, unbalances the plot with a cluster of episodes all tilted in the same direction. Ideally, in the long run there should be reciprocity, and so, at the scale of the novel as a whole, and symmetry should be restored: but at the scale of the chapter, a-symmetry is exactly what a narrative of guanxi should be like. And, needless to say, a novel which is a-symmetrical at one scale, and symmetrical at the higher one – this is an amazing construction.
FRUITFUL DOING... So far, I have looked at how individual behaviour contributes to the shape of the network; now I'll turn the matter around, and see how the system of relations shapes the individual character. Bao-yu, in chapter 8, is a good instance (fig. 53): as the narrative unfolds, he takes part in at least three distinct episodes: he has a particularly important encounter with his pre-destined bride Bao-chai, which is catalysed by her maid Oriole; then he gets drunk, despite Nannie Li's vigilance, amidst all sorts of banter from the characters around him; finally, he throws a tantrum with his maids, until Aroma threatens a general desertion. Three episodes; three totally different sides of Bao-yu: naive lover, sensuous youth, petty domestic tyrant. And something similar happens at every chapter: the novel's huge pack of characters is re-shuffled, and the new "hand" brings out new groups, and hence new features in the figures we already knew. Novelty, as re-combination: in twenty chapters, Bao-yu speaks to 54 characters, and not once does the same group re-form around him.
Unlike the first two characters I have discussed, who are both definitely minor figures, Bao-yu is arguably the protagonist of The Story of the Stone: the male child born under very special auspices, and expected to do great things for his family. But he leads a strange life, for a protagonist: constantly summoned by this and that relative, kept under supervision, asked to perform all sorts of duties – even the many delightful opportunities he is offered come usually from above, and with constraints attached. He is the protagonist, yes, but he is not free. He is the protagonist, and therefore he isn't free: he has a duty towards the structure: towards the relation-based society he is part of. "The One for the Many", rather than "versus": Elizabeth Bennet, not off to Pemberley on her own, but kept at home, to shape the life of her sisters.
It's an issue that had already come up: network theory, and Woloch's theory of character. And one thing is clear: network theory fully corroborates Woloch's concepts of character-space and character-system: they were a bold intuition, and have now a solid empirical basis. The protagonist is a trickier case, since these first networks suggest more of a continuity between the One and the Many than a categorical break: the "strong" protagonist – the one indispensable to the existence of the network: Hamlet – being so far the exception, rather than the rule. Besides, the notion of "interiority" – indispensable to Woloch's idea of the protagonist – is not pertinent within network theory: not pertinent, because this theory works on interactions between vertices, whereas interiority unfolds within a single vertex, at a textual level which is vertical, rather than horizontal. Paradigmatic, not syntagmatic. In this sense, the encounter of the two theories could lead to a stricter distinction of levels within Woloch's general model.
Which leads me to my final point. In general, what relationship ought to exist between a literary theory, and models drawn from other disciplines? Aside from a pious academic shibboleth, what do we really want from "interdisciplinary" work? Something new? Of course. But new how, vis-à-vis existing literary theories? A paradigm shift? That's always exciting; but, it usually presupposes a field trapped in Ptolemaic epicycles. Is that where we are? Some think so: the darwinian collection The Literary Animal, for instance, quotes around seven hundred works, and among them there is nothing by Adorno, Auerbach, Bakhtin, Barthes, Benjamin, Benveniste, Brik, Curtius, Eco, Frye, Jakobson, Jameson, Lukacs, Shklovsky, Spitzer, Todorov, Tomashevsky, Tynianov, or Weinrich. They do quote Propp once in a footnote, yes, but one feels that for them a century of literary theory has been a perfect waste of time. I disagree; and think that what we should look for, in an external model, is not so much a new dawn for literary studies, as a broader framework for those theories that already work well. People are often surprised by the fact that I study evolution, cartography, statistics, factor analysis, networks ... and then keep quoting Shklovsky and Barthes and Auerbach all the same. But they miss the point. Evolution is a fantastic theory; but I became convinced it would work for literature only when I realized that it was consistent with all the key concepts of Russian Formalism, and a host of solid historical findings from other sources. Evolution was a good framework, because it agreed with what had been found independently of it.
A framework that agrees with previous findings ... That's all? Nothing new? No. What is new, is the agreement. Corroboration. Showing that some theories really work, and can be part of a wider system. For science, Stephen Jay Gould once wrote, fruitful doing matters more than clever thinking. One day, it'd be good to be able to say the same about literary studies.
Franco Moretti
SvD, 29 November 2009
- Conferences > NEWW November meetings > 2009 > Network Theory