Friday, January 13, 2006
Book Notes: Franco Moretti’s Graphs, Maps, Trees
Of all the odd things I’ve heard in recent years, one of the oddest would be that there are objections in principle to the research paradigm that Franco Moretti describes in Graphs, Maps, Trees. It really doesn’t matter what your interest in cultural or literary analysis is: what Moretti proposes is useful grist for your mill. There is no requirement to purchase the entire methodological inventory he makes available, or to throw overboard close reading or aesthetic appreciation or focus on a small and rarified set of texts. Frankly, when academics propose that we only do what they’re doing and stop doing everything else, I tend to ignore such propositions in the same way that I ignore commercial hyperbole while deciding what things I want to buy. I enjoy my iPod: I’m not required to think that it has changed my life or should lead me to chuck my stereo out the window. Whatever you think literary analysis and cultural history are, quantifying the subject of their domains is a very good thing. Indeed, it is a kind of knowledge long inferred and rarely acquired, and though its acquisition unsettles some assumptions made in the inferred known, it equally clarifies and strengthens many other claims--or least puts new and productive burdens on them.
Leave aside for the moment the particular kinds of modellings and configurations of his data that Moretti describes, and just stick with the numbers alone. Even in a single national literature, it used to be hard to make any clear statements about the total number of books published in a given year or across a long series of years, and of those books, what proportion were works commonly known, analyzed, or regarded as defining a “literature.” Now Moretti is not really so unusual or isolated as he might appear in taking an interest in such quantification, as Matt Greenfield has already noted. There are many subfields of cultural history and literary analysis that have taken an interest in similar quantification and mapping, in fact, the study of genres has long been shaped by an interest in cycles of publication of the kind Moretti describes.
The numbers alone, as Moretti observes, immediately falsify or complicate a series of conventional ways of understanding cultural or literary change over time. When we speak of a particular novel’s influence, or about how literature changed in response to a particular work, we’re making claims that ought to involve a total topography of published cultural work. Until recently, that would not have been the case. If it turns out that that the lineal descendents of a novel regarded as influential are no more than half a percent of all work published over a ten-year period, this puts pressure on what we mean by “influential”. It is not that we are now forbidden to make the claim, but it constrains and specifies what we can potentially mean by such a claim. It’s just that what Moretti does helps us to realize that often in making such claims, we’ve put too much trust in the representations and attributions of authors and readers, which are just as produced and fantastical as any publically uttered memories, just as Goffmanesque in their performance as any other presentation of self. It is not that we are forbidden either to speak of that novel’s quality or desirability, of what we (and past readers) might have found enticing, inspiring, productive, mysterious in such a work. Moretti doesn’t quantify the production of meaning, and even if he wanted to, he could not.
Enough on the simple virtues of Moretti’s project. Of course cultural historians and literary critics need numbers, all of us, and godspeed to the counting and graphing. I’d love to see someone do something similar with major historical archives: count all the documents, all of them, and graph for me their types and forms. Historians live in their archives, but we don’t really know them half as well as we ought to. We accept the categories that the archive offers us, and read along the pathways laid down. In researching consumerism and material culture in colonial Zimbabwe, I had to read horizontally across an archive for a topic that the archive itself did not recognize as lying within its confines, and the sense I got of what the archive contained was thereby complicated very much from what I’d thought it to be. Quantification could only help that understanding further.
What could enhance Moretti’s work further? What do I see as genuine problems and gaps in the models he offers?
First, a warning: that counting publications only scratches the surface of the totality of cultural production in any given post-Gutenberg moment. This is an issue that Raphael Samuel wrote about for years with regard to historians and their archives: that what lands in archives, is recorded as documentary evidence, is just a small and sometimes highly unrepresentative selection of the totality of potential grist for the historian’s mill in a given era. Moretti may be counting formal publication and finding that what is commonly taken to represent “national literature” is not typical or representative, but beyond that lies an even larger domain composed of the ephemeral, the unpreserved, the unrecorded. In the age of electronic communication, we should be especially wise to this problem. Even with the Web being archived, much of what has been written within it and read avidly is likely to be lost in the longer-term: asynchronous discussions, epistolary literatures passing through email, and so on.
There will come a point at which a project of quantifying cultural production in any given historical moment will only be able to gesture at a vast Oort cloud of unknown writings, performances, and texts, seeing the gravitational effects of some unseeable and lost Planet X tugging at the knowable and quantified. This especially strikes me as an Africanist: we now have some lovely examples of “market literature” in Nigeria available in published form, but beyond those examples, I very much doubt we will ever be able to represent the numbers or varieties of such texts published. If we confine our understanding of what was typical or normal within a cultural form to what we can find in archives, in libraries, in catalogs, in records of publication, we’ll ultimately have a deformed conception of the totality. Beyond everything counted there is always another mountain of the uncountable. Historians of slavery turned over every stone and record to count the total numbers of Africans taken across the Atlantic, and even then, had to make some educated guesses, which still fuels (sometimes quite intense) debate among specialists in that field. But once some numbers were in hand, those historians realized that making any statements about their meaning depended on another set of numbers, namely, how many people there were in West and Equatorial Africa at any given moment in any given society, of what the fertility rates were in those places, of the numbers of men and women, and so on. All numbers which, frankly, are never going to be tallied through anything besides serious guesswork.
The second thing that occurs to me on reading Moretti is that we know quantifying publication and quantifying discrete elements (tropes, places, and so on) within publications doesn’t tell us half so much as we might think about the quantification of readership and circulation. Again, maybe it’s because I’m an Africanist that I’m especially wary in this regard. You can count up the numbers of newspapers published in a decade in southern Africa, including ones presumptively aimed at African audiences. You would be making a big mistake to assume that such numbers tell you how many people were reading or consuming those newspapers. We know from historical and ethnographic work that the literate often read or reinterpreted newspapers for the illiterate, and that a single copy of a publication was often passed around many readers. Texts travel through readerships in ways that numbers do not describe very well. Here I’d look to Isabel Hofmeyr’s fantastic book on the transnational history of John Bunyan’s Pilgrim’s Progress for some insight, for a tracing of how a single work can traverse readerships in ways not precisely correlated with its appearance in libraries, archives, or even within texts that invoke, allude or cite Bunyan. There ought to be a sociology and social history of audience and reading that might complement Moretti’s work, but my intuitive suspicion is that it would also very much complicate the claims he would like to make. I also think that the sociology of authorship and publication would be a useful complement to Moretti: to know who knows whom, who reads whom, and to which outlets and forms of publication they relate strikes me as retaining its importance.
The most important concern I have about Moretti is that I think he has the same problem that the Annalistes and world-systems analysts have had with modernity: a difficulty explaining rupture, breach, or novelty. Novelty here in multiple sense: as Elif Batuman observes, the novel-form is what gets marked off in Moretti as something not explained. In world-systems history, this problem has lately been exaggerated to extremes by some of the founding practitioners in the field, as in Andre Gunder Frank’s argument late in his life that the contemporary world-system is part of a continuous five-thousand year old history, that modernity or the rise of the West is a temporary or epiphenomenal speed bump in a well-worn road, not anything genuinely new. The problem with a divergent tree of literary or cultural history is that it has a hard time explaining the appearance of genuinely new forms or genres: it is forced always to insist on a fundamental continuity. The best that the world-systems historians could do, if they didn’t want to follow Frank’s argument that modernity or the rise of the West was an illusion, was to either insist on materialist explanations of rupture (new technologies, new means of production) or to offer shopworn dialectics.
In evolutionary terms, Moretti is something of a gradualist; my impulse is to throw up the cultural equivalent of punctuated equilibria in reply, to insist that some genres and forms do not descend gracefully from predicates but emerge abruptly, catastrophically, like Aphrodite stepping from the waves. The evolutionary metaphor is a powerful one, but you want to take in even more of it than Moretti does. For one, it’s fine to talk about the death of forms and genres, about how divergence fuels convergence that fuels more divergence. You can’t have a metaphor that invokes evolution or speciation without death, or at least the removal of specialized forms. But it begs the question (and Moretti knows that it does) of what the fitness landscape is for cultural forms.
Emerge in fact is the operative verb here: I think Moretti’s trees in particular could benefit enormously from reference to the body of work subsumed under the heading of “emergence” or “complexity theory”. Because there is an answer within that body of work to Moretti’s question: what explains the divergence of literary forms. It’s not an especially comforting answer, perhaps, for either Moretti or some of his critics, because it may eschew some deep underlying explanatory principle for why some genres, tropes, modes of literary representation produce an explosion of divergent forms and why others die. In an emergent system, the place within the topology of the system where complex structures appear may be effectively random. If we take Moretti’s example of Sherlock Holmes, it might be that an evolutionary tree of British fiction in the last half of the 19th Century would help us to understand why the environment was friendly to “detective fiction”, what the conditions of the cultural soil were like for the growing of a new tree. But as for how Doyle’s stories set the conventions of a genre and others die, are forgotten, wither, some of that might be simply termed “dumb luck”, that the precise location as which crystallization of a genre occurs in a moment where many nascent forms of the genre are present is about the accidents of readership, of circulation, of publication, of imitation, that there is no deeper explanation that needs to cite how Doyle’s particular formulation of the genre more precisely satisfied or represented the desires of a reading public, or how his ability as a writer was more precisely distinguished from any other. I’m echoing Gould’s Wonderful Life here very consciously. This is a rebuke of traditional literary theory, historicist literary theory and even Moretti all at once: all of them assume that there is a rational way to explain cultural reproduction which relates the successful, generative or meaningful text to some underlying condition of its being: an ideological or discursive fit to its environment, a skillful or superior authorial creation of an aesthetic, or some undiscovered underlying “law” of cycles and divergences. Here maybe Moretti needs to go the next step rather than running back for the materialist security blanket as he does in closing the book.
The accidental and the emergent are also, however, where we might reopen the door to agency, creativity and the will of the author and reader again. Because another thing that appears in literary and cultural history is the unpredictable generativity of authors and readers who reach from a high branch far back down the tree to create some new possibility of representation, who take what was a junk gene in DNA of culture and from it express some meaning or representation that was deemed impossible the day before. Sometimes such authors are just Carlo Ginzberg’s Menocchio, envisioning private cultural worlds that die or are forgotten; sometimes they are better situated, differently located, or even, dare we say it, more imaginative or skillful in how they excavate the literary past in order to produce new possibility. Just as I would in the end say that modernity is an emergent and in some ways accidental social structure which in turn creates the possibility for individual agency that then generates still other emergent forms through will, choice or deliberate selection, I think you can reconcile the agency of authors and readers with Moretti’s graphs, maps and trees, but it does take coloring outside his lines to do so.
>Because another thing that appears in literary and cultural history is the unpredictable generativity of authors and readers who reach from a high branch far back down the tree to create some new possibility of representation.
This is part of what makes the kind of trees Moretti wants to do so hard to do. Every author can be influenced by a huge number of previous authors which makes the construction of accurate trees very difficult.
>But as for how Doyle’s stories set the conventions of a genre and others die, are forgotten, wither, some of that might be simply termed “dumb luck”
I think dumb luck can be a factor, but I think, if anything, Moretti over estimates the randonmess in genre formation.
when mystery writers come up with an ‘aeronaut’ who kills a hiker with the anchor of his balloon, or a somnambulist painter who draws the face of the man he has murdered, or a chair that catapults its occupants into a neighboring park, they are clearly looking for the Great Idea that will seal their success. And yet, just as clearly, aeronauts and catapults are totally random attempts at innova-tion, in the sense in which evolutionary theory uses the term: they show no foreknowledge—no idea, really—of what may be good for literary survival. In making writers branch out in every direction, then, the mar-ket also pushes them into all sorts of crazy blind alleys; and divergence becomes indeed, as Darwin had seen, inseparable from extinction.
Unlike purely random mutations in biological systems, authors can use their judgement to pick a plot that they think will work well. They aren’t always right, of course
Right. Which is another pet interest of mine that I think Moretti’s methodology can actually help illuminate: the history of failure in cultural production. I see this for example really acutely in the historiography of work on advertising: most scholars begin from the point of trying to explain the success or productivity of a given advertisement or advertising campaign, and don’t consider at all the significant number of failed campaigns, failed advertisements or failed products, which is actually a necessary accompaniment to any claims about why a given advertisement succeeded or reproduced itself in some fashion. I’d say that’s true about almost any text or genre of texts.
They aren’t always right, of course
No, the’re not. And that’s one thing that’s so interesting. The process of writing and publishing books is jam-packed with deliberation and intention. The author wants to tell a certain story, the editor wants to make the text better, the publisher wants to sell lots of copies. And yet the book’s fate in the marketplace is deeply mysterious. The books that do poorly are as deep in intention and deliberation as those that do well.
The most sophisticated treatment of this sort of thing that I’m aware of concerns the movie business, Arthur de Vany’s Hollywood Economics: How extreme uncertainty shapes the film industry:
Timothy: Thanks for your defense of Moretti’s methodology. The concerns you deflate are so often tossed at “the sociology of literature” in general: it’s the end of close reading, critics don’t actually have to read books, critics are just looking at covers or blurbs, etc.
Still, I think the literary public is becoming more and more interested in what these methods can teach us. All of which is to preface this link to Louis Menand’s positive *New Yorker* review of James English’s *The Economy of Prestige* and Pascale Casanova’s *The World Republic of Letters*, two important works in the sociology of literature. English and Casanova aren’t as quantitative as Moretti, but each is looking at a much wider spectrum of literary phenomena than most critics.
Literature as case study or data seems mostly misguided. Will data mining on say Kafka produce something even remotely connected with the text? The Morettian analysis appears to be concerned with the economics of writing and publishing, rather than lit., which is not necessarily a wrong tactic, but then why not simply examine the various big publishing houses, and do some sort of analysis regarding the economic demographics and data........do lit. scholars now track fluctuations in Random House stock to make assessments about literature.........? weirdness
Will data mining on say Kafka produce something even remotely connected with the text? The Morettian analysis appears to be concerned with the economics of writing and publishing, rather than lit., which is not necessarily a wrong tactic, but then why not simply examine the various big publishing houses, and do some sort of analysis regarding the economic demographics and data........do lit. scholars now track fluctuations in Random House stock to make assessments about literature.........? weirdness
The question isn’t do lit. scholars track fluctuations in markets, but should they. Does literature somehow float in a space that is entirely independent of the economics of writing and publishing? Are authors’ choices of themes, settings and characters unrelated to their beliefs about what will sell in the publishing marketplace? I’d suggest that these questions answer themselves.
And it’s not as though those market forces are independent of readers interests and tastes. On the contrary, the logic of de Vany’s argument is that readers drive the market, not vice versa. No matter how hard movie makers and publishers and recording companies attempt to manipulate and control their marketplaces, they fail. Hence those businesses are volatile ones, with firms rising and dying by the decade.
“Literature as case study or data seems mostly misguided."
The journalistic take on all this describes quantitative methods as if they’re in opposition to reading, and Moretti’s tone has encouraged that take to some degree.
But as Moretti, Burke, Greenfield, and I have all noted on this site, it isn’t (or shouldn’t be) made into an either-or choice. Insofar as there’s any overlap, such studies clear a space for literary reading. To put it another way, their findings sometimes make it less likely that we readers will prattle nonsense. (Assuming that the researchers don’t start prattling nonsense themselves....)
I am in favor of quantitative methods, but that doesn’t mean one is in favor of the methods used by Dr. Moretti (honestly I am not sufficiently acquainted with his research to offer a fully informed assessment). Not many people would would question the value of basic social psych. experiments such as Milgram studies on obedience, or Tversky’s research challenging the “rational man” standard. Yet literary texts are not really any sort of observable behavior, nor really the sort of syntactic or morphological evidence linguists examine, are they? As far as data mining providing information leading to what the “real” meaning of the text is, that seems as likely to produce the truth as a consumer poll regarding what Americans think of, eh, Macbeth and then offering the most popular or consistent interpretation as the meaning.
But the data mining folks are suggesting that they’re trying to arrive at the “real” meaning of the text. While some evolutionary psychology types talk about texts having determinate meaning (and ev psych is the way to it), I don’t think that’s a common belief or aspiration. And Ellen Spolsky has argued that it is precisely because textual meaning is elastic that literary texts are so culturally useful. If the meaning of texts were rigidly fixed, then they’d be trapped in their historical moment.
Spolsky, E. (2002). “Darwin and Derrida: Cognitive Literary Theory As a Species of Post-Structuralism.” Poetics Today 23(1): 43-62.
Whoops. “But the data mining folks are NOT suggesting . . . “
I agree with Ray that Moretti’s methods can produce information that is helpful to even the most literary of close readers (like me, for example).
I also agree with Ray that projects like Moretti’s “sometimes make it less likely that we readers will prattle nonsense.” But at some point the consideration of literary works must indeed “float in a space that is entirely independent of the economics of writing and publishing” or it’s no longer a consideration of *literature*. It’s economics or history or whatever. What aesthetes like myself object to is not per se approaches to literature that use it to explore cultural history but the fact that too often the study of literature is reduced to *only* such approaches, which leaves literature indeed merely “culturally useful.”
Doesn’t the contingency argument fail on cultural evolution because of the sensate and consciously malleable substrate? If there are no atomistic units of cultural heredity (do even memeticists argue for atomicity?), then “dumb luck” emergence must operate at a much higher level of abstraction, which is compatible, I think, with what you’re saying.
I believe that some memeticists do, in fact, argue for atoms; we’ve got memes, and memeplexes and other meme thingies. But some of them do not. And there are some scholars who explicitly argue that we do not need discrete units:
Common misunderstandings of memes (and genes):
The promise and the limits of the genetic analogy to cultural transmission processes
Francisco J. Gil-White
Abstract: Memetics suffers from conceptual confusion and not enough empirical work. This paper will argue that, although memes are not, in fact, selfish replicators, they can and should be analyzed with Darwinian models. The selfish meme conception does more to distort than enlighten our understanding of cultural processes.
There is now a vigorous debate on how Darwinism should be applied to culture (see Aunger 2000). Following Dawkins (1989), many now refer to units of cultural transmission and evolution as ‘memes’, regard ‘replicators’ as essential for a Darwinian process, assume ‘selfish memes’, and adopt a ‘meme’s eye view’.
Analogies and borrowed yardsticks are often useful for a new field, but may also cause misunderstanding. I will argue that Dawkins’ legacy for cultural Darwinism has not only given rise to confusion but itself results from misconstruals of Darwinian theory.
I will not define a ‘meme’ as a selfish replicator 1 but will adopt the broad Oxford English Dictionary’s definition—“an element of culture that may be considered to be passed down by non-genetic means”.2 Selfish replication, then, is a hypothesis about the behavior of the stuff that gets transmitted through non-genetic means. The relevant questions, then, are:
1) Does this stuff behave like a selfish replicator?;
2) If not, does this make Darwinian analyses of culture impossible?
3) If not, is it impossible to find the boundaries of memes?
4) Can we simply appropriate the ‘selfish gene’ idea from biology?
I will answer “no” to each but will still call what is culturally transmitted ‘memes’.