James F. English
FRANCO MORETTI RELISHES HIS IMAGE as a maverick and a troublemaker. Back in 2000, when he first he proposed distant reading as a method (in his essay “Conjectures on World Literature”), he could have traced a respectable pedigree among other historians of the book who work on large and extra-canonical corpora, such as Robert Darnton, William St Clair, or John Sutherland. Instead, he took up a position beyond the pale, casting aspersions on critics who engage in close study of individual works, dismissing that standard approach as “theological” and urging its adherents to rebel. What he offered to the discipline of literary studies was a “pact with the devil”: “we know how to read texts, now let’s learn how not to read them.”
Such utterances have led to polemics of a rather tiresome kind, unhelpful both to book history and to the digital humanities. But if the controversies surrounding Moretti seem a bit bogus, the experience of reading him is never less than exhilarating. No one conveys better the sheer fun of literary research, or invests the discipline with such an air of discovery and possibility. In a field that has become dour and down on itself, Moretti stands out as an unflagging optimist.
The writer Elif Batuman, who studied with Moretti at Stanford, describes his work as lending to the science of fiction a “science-fiction thrill.” And indeed, the sequence of essays collected here reads a bit like an SF adventure story, each unfolding phase of the action bringing with it some new gadget or technology of reading: the evolutionary tree, the world-system atlas, the character-network map. The pleasures of detective fiction are on offer, as well. Not only because, like any historian, Moretti is unearthing in the present clues about what happened in the past, but because the clues he sets before us are so unlikely and so inscrutable, yet yield so readily to his analytic powers. Having performed a count of definite and indefinite articles in the titles of various genres of 19th-century fiction, Moretti discovers that the definite article predominates in titles of anti-Jacobin novels (outnumbering indefinite articles 36 percent to 3 percent), while the indefinite is more common in titles of New Woman novels, appearing there 10 times as often. Why should this be? Elementary, my dear Watson. The linguistic function of articles is to direct a reader’s attention either backward, toward the already established, or forward, toward a new and unknown thing. The anti-Jacobin novel depends on received ideas, the New Woman novel on an encounter with something unprecedented. Moretti is ever alert for such puzzling quirks in the data, and never at a loss for the nifty Holmesian explanation.
But more than science fiction or the detective story, Distant Reading summons to mind the picaresque. Already when they appeared as journal articles (most of them in New Left Review), these essays were notable for “showing their work,” scratch-outs and all. As they are mostly essays on method, it makes sense that they would be more about process than outcome. But Moretti’s style involves a particular eagerness to highlight his own false steps and to show us the dead alleys down which they have led him. It is not simply a trick, to inflate our yield of pleasure and admiration when he at last locates some bright new boulevard of historical knowledge. Far more often, we find Moretti striding along the grand boulevard, praising its attractions, until some stranger comes along and points out that it’s just another cul-de-sac after all.
So, for example, “The Slaughterhouse of Literature,” the essay on clues, which Moretti had intended to illustrate “the method for the study of noncanonical literature,” ends with an admission that it is actually a one-off, leading nowhere. The essay had shown rather brilliantly how the Darwinian literary marketplace of Victorian Britain selected for detective stories with properly working clues, enabling the genre to emerge in its recognizable form while pushing many inferior variants (stories whose clues were not necessary, visible, and decodable) into extinction. But the whole project, as a Columbia graduate student first pointed out to him, amounts to an exercise in tautology: Moretti had, in effect, proposed a method for reading beyond the canon that is “so focused on a canonical device […] that in the noncanonical universe it can only discover […] the absence of the device, that is, of the canon.” What such a method has to teach us about literary history is thus “true, but trivial.”
Moretti’s new headnotes push this trash-binning tendency even further, to the point where very few of the models and concepts he has constructed over the past decade and a half are left intact. The entire Darwinian model of literary history, which he developed in “Modern European Literature,” “The Slaughterhouse of Literature,” and “World-Systems Analysis, Evolutionary Theory, Weltliteratur,” is dispatched in a headnote to the last of these. The trouble with the Darwinian model is that “evolution has no equivalent for the concept of social conflict,” that is, for the kind of conflict that can “redefine the entire ecosystem.” This is no incidental defect, but a fatal blind spot “which, clearly no theory of culture and society can allow.” It is also, as Moretti points out, a blind spot shared by network theory, so another of his favored models for literary study perhaps bites the dust here as well. A footnote to the same essay concedes that the theory of world literature it advances likewise needs to be discarded or “completely reformulate[ed].”
This is presumably because (as detailed in yet another headnote) there turns out to be a fundamental flaw in the literary world-systems model that provided the conceptual backbone for such pieces as “Conjectures,” “More Conjectures,” and “Planet Hollywood.” The economic world system as developed by the sociologist Immanuel Wallerstein is defined by an interdependency of its constituent elements, with the core just as dependent on the periphery’s cheap unskilled labor as the periphery is on the core’s technology and capital. Moretti’s approach has assumed a homology between the world-economic and world-literary systems. The problem, he now concedes, is that, in his model of world literature, all vectors of explanation proceed in one direction only, allowing for dependency but not interdependency: “the activity of the literary core [which invents and globalizes major forms such as the novel] largely determines developments in the periphery, but the reverse is usually not the case.” He is thus really working with only half a world-systems model; and Moretti, suspecting that “half a model […] is no model at all,” concludes, “I should have known better.”
Distant Reading is picaresque, then, in detailing, via a witty and roguish, foreign-born, first-person narrator, a series of (mis)adventures in which things can be counted on to go wrong, but never so wrong as to deter the protagonist from embarking on further adventures. Like any good picaresque hero, Moretti takes his setbacks in stride, and never gives up on the idea of progress. The very fact that his ingenious models can be knocked to pieces, and his hard-won results summarily falsified, serves as a kind of negative validation, a sign that he is on the right track, practicing a genuine science of literary form, anchored — unlike the softer, safer, merely interpretative work of more tractable critics — to the bedrock of empirical evidence. For all his ready abandonment of specific scientific models, and his “aware[ness] of how few concrete results have emerged” from them, he has never retreated from his positive embrace of science as such, or his simple faith in quantitative method, raw data, objective facts. He has held firmly to the view, expressed in 2007’s Graphs, Maps, and Trees, that empirical data are the key to better knowledge of literature because, while we must ultimately interpret them, the data themselves are, or should be, “independent of interpretation.”
There is something old fashioned, even innocent, about this faith in data. Naïve ontologies of the dataset — equating quantification with objectivity, data with facts, and bigger data with better access to reality — have been thoroughly critiqued by historians and sociologists of science and, more lately, by digital humanists. (Witness the fine recent collection of DH essays edited by Lisa Gitelman, Raw Data Is an Oxymoron.) From the standpoint of such critiques, every kind of literary data is, by definition, already cooked according to some recipe of literariness, which is in turn derived from some history of reading and, yes, of interpretation. However annoyingly half-baked our concepts, there really is no escape from the kitchen. Our chefs have already had a hand in determining which ingredients will be found in which market stalls; how they will be sorted and arrayed and valued; and who will grow them where. The buoyant optimism that propels Moretti from one unlikely literary experiment to the next seems to depend on averting his attention from these ontological matters and maintaining an uncomplicated enthusiasm for data-driven research.
It is a strategic innocence, perhaps, as well as an enabling one. But it leaves Moretti, at the end of this collection, putting the question of distant reading’s future in too one-sided a form, asking only whether the rest of us will ever learn how to do properly scientific work, whether we are “capable of sharing raw materials, evidence — facts — with each other” in sufficiently organized fashion to gather the “enormous amount of empirical data” that is required. Given the institutional upheaval involved in shifting humanist scholars, so habituated to working alone on low but steady budgets, onto the laboratory model of research, with its large, expensive teams and nonstop grant-writing, Moretti is right to wonder. But he should be asking, as well, whether he and other digital pioneers will be capable of demonstrating, to those who are more skeptical of the relations between materials, data, evidence, and facts, that such a massive upheaval is worth the effort.
Alexander R. Galloway
Everything Is Computational
In the mid-1980s, the media theorist Friedrich Kittler, an émigré from East to West Germany proficient in literature and philosophy, came to a provocative conclusion. After conducting deep histories of media and analyses of both literature and philosophy, Kittler made a claim that would scandalize many: all of literary theory is media theory. Friedrich Nietzsche's writings may be interesting, but more interesting to Kittler was the typewriter that Nietzsche used to compose them. The novel Dracula may be a masterpiece of gothic horror, but for Kittler it was most significant as an orgy of media formats. The great works of literature and art were thus also great works of media, both thematically, insofar as they represented media technologies within their narratives, and physically, as tangible media artifacts themselves.
At about the same time, in Paris, the philosopher François Laruelle had started to publish his first systematic analyses of philosophy as a whole. Like Kittler, he came to a startling conclusion: all of philosophy is media theory. Unlike Kittler, Laruelle had little interest in writing machines and phoneme generators. He found nothing particularly fascinating in Nietzsche's typewriter, nor for that matter in Martin Heidegger's radio, or Jacques Derrida's Xerox machine. Instead, Laruelle asserted that philosophy itself was an act of mediation, for the most elemental philosophical stance was that of reflecting about and interacting with a world. He derisively labeled philosophers the mere “mailmen of truth,” their profession a glorified media infrastructure for minds and bodies, essences and entities. For Laruelle, Hegel's dialectic was little more than a grand machine for mediating history, Deleuze's plane of immanence little more than a giant sandbox for multiplicities to intermediate.
Like that of Laruelle and Kittler before him, the work of Franco Moretti indicates an important historical shift in the humanities and the way it defines its task. It is not simply the meaning of literature that must be pursued but the fact of literature, its bare existence as a technical object. Moretti is not the first to pursue this idea. Just as one might read and analyze Sigmund Freud's The Interpretation of Dreams, one might also read and analyze Freud's mystic writing-pad (as Derrida did). And just as one might study Derrida's Writing and Difference, one might also study his mediatic performance (as Avital Ronell did). Indeed, by the end of the 20th century, a whole series of thinkers had come to the same conclusion, that both literature and philosophy — and many other things too — had to be understood as systems of mediation. And, being machines themselves, such systems could, in theory, be read by other machines. They could be scanned, parsed, processed, and recombined into new knowledge.
Neither Kittler nor Laruelle set out to rearrange the grand hierarchy of the disciplines. Media systems are not more fundamental or more capacious than literature and philosophy, as quantum mechanics is more fundamental than chemistry. Media are not merely receptacles to contain literature and philosophy, as a piece of paper may contain the written word. Rather, their contention is that both literature and philosophy are already media theory, and that reading, interpretation, reflection, analysis, and synthesis are already processes of mediation. Thus, to reflect on arts and ideas, as Plato did in the Phaedrus and Kant did in the third Critique and Marshall McLuhan did in The Mechanical Bride, is to reflect on the very techniques of mediation. Indeed, as Laruelle contends, the most emblematic mode of mediation is the simple desire to reflect in the first place.
What would it mean to say that the center of thought and culture is not literature, art, or philosophy, but media? This seems to be Moretti's provocation in a nutshell.
So let's dispense with the silly questions up front. Does the kind of macroanalysis described in Distant Reading produce useful results? Should scholars diagram the plot structure of Hamlet? Will it help to database all of Victorian literature for parsing by the computer? Do clustering algorithms work? Should digital methods be used in the humanities?
These kinds of questions make certain people nervous. Some humanists have defended what they see as the incommensurability of culture against the profanations of evolutionary models (Christopher Prendergast writing in the New Left Review); others bemoan the loss of literary value, as canonical masterpieces are compared side-by-side with reams of derivative pulp (Stephen Marche writing in the Los Angeles Review of Books). And some Luddites, be they proud or private, feel lost or left behind by the computer revolution, preferring to comment ironically on the increased industrialization of contemporary life, or reject it completely. (I'm caricaturing, of course, and I'll admit to being seduced, at least partially and sometimes against better sense, by aspects of several of these factions.)
To be sure, Moretti's system has the potential to produce interesting results, as his own books attest. Yet a deeper issue often gets missed in the hullabaloo around his trademark “distant reading.” Couldn't we have seen this coming? The computer arrives on the scene, and presto — everything is computational. Silicon Valley transforms the economy using clustering algorithms and, a few years later, a Palo Alto professor starts proselytizing about how such algorithms can be used to produce valid literary evidence. We have had our champions of network-centric warfare (Donald Rumsfeld) and network-centric capitalism (Larry Page). So why not network-centric sociology (Bruno Latour) or network-centric literary criticism (Moretti)?
Shouldn’t we challenge this kind of techno-mimicry when it occurs, not from a traditional humanist point of view, but from a political one? Why recapitulate the technical infrastructure of late capitalism in our scholarship? And in the wake of such recapitulation — here is the point that most confounds me — why should Moretti continue to trumpet his endeavor as Marxist? Wouldn't blindness to the technical infrastructure be evidence, instead, of an anti-Marxism? Similarly, in an age when the means of production are so difficult to see and understand, don’t we need a stronger hermeneutics, not a weaker one? More interpretation, not less?
Recently I asked a friend what she thought about “digital humanities” and, more generally, the current trend in academia away from interpretive criticism and toward quantitative or cognitive research methods. “Digital humanities?” she replied, her face knotted with skepticism. “Isn't that for old professors who don't understand computers, and young ones who don't understand hermeneutics?”
While I don't entirely share her skepticism, I took her point. Indeed, as scholars, we are at the beginning of an entirely new phase in the history of the humanities. We have only begun to understand what it means to view art, literature, and culture as fully technical and mediatic systems. Yet, in the long run, the work of media archeologists like Kittler seems more promising to me than Moretti’s, since it provides a means for doing research without obfuscating those means, as Moretti does. Many scholars and students who perform data mining do not entirely understand how these algorithms work, and those who do still often naturalize such technologies as merely neutral tools for empirical fact finding. (As for Laruelle, his utility for any kind of literary criticism remains to be seen.)
So I recommend you skip Distant Reading and read The Bourgeois: Between History and Literature, the second of two Moretti books published this season, instead. A perfect antidote to the tedious academic squabbles chronicled in the first book, The Bourgeois offers a glorious mental ballet, a testament to Moretti's virtuosic style as a thinker and critic. Can he write! And practically no bean counting in this one — no graphs, no maps, no trees. So are there two Morettis? Two dueling methodologies? In an aside in Distant Reading, he concedes the strange incompatibility of the two books, but withholds further explanation: “The fact that Distant Reading is published alongside The Bourgeois — a book that couldn't be more unlike it in spirit and execution — makes me think that I prefer studying tides and moon independently of each other.”
The techno-mimic or the literary lunatic? For me, an easy choice.
The Ends of Big Data
On or around June 6, 2013, it became impossible to consider writing about harnessing the power of Big Data for the humanistic enterprise without contemplating its power for more distinctly inhuman enterprises. Of course, the two have gotten conflated — suddenly people who'd never before heard of metadata are frightened by it, without knowing much about the more useful and benevolent purposes to which metadata can be put — but what the PRISM leaks have made clear is that computational approaches to understanding vast quantities of information have the potential to disrupt and transform our understanding of the culture in which we live.
This brings me, perversely, to Franco Moretti's Distant Reading, which gathers essays that provide an overview of the evolution of Moretti's somewhat controversial methodologies over a 20-year span. As Moretti notes, dealing with large quantities of data requires a very different set of practices than traditional literary scholarship. “[O]ne cannot study a large archive in the same way one studies a text,” he writes:
texts are designed to “speak” to us, and so, provided we know how to listen, they always end up telling us something; but archives are not messages that were meant to address us, and so they say absolutely nothing until one asks the right question.
Archives are not messages that were meant to address us, and yet meaningful information can be extracted from them. This is an exciting idea when the archive is the entirety of literary history, a chilling one when it's our private internet activity or phone records.
Given the retrospective tone of the present-day introductory notes with which he prefaces each essay, one begins to suspect that the volume’s title is meant to hint at a kind of distant reading of Moretti, by Moretti. But given that the distance involved is one of temporality rather than scope, and that Moretti does not present any computational analysis of his work at scale, the typical reader’s engagement with Distant Reading is, instead, very close. We are thus led to focus not just on how Moretti reads, but on what he reads in that distinctive way, and even more importantly why he does this — a cluster of distinctions that recent public discussions of data mining and surveillance have somewhat elided.
What Moretti reads, of course, is the history of national literatures as they have evolved, differentiated, and cross-pollinated over the last several centuries. Why does he read this history distantly? Why does he treat that literature as data, rather than as representations? In order to understand literature at a different scale. Rather than exploring the relationships between this novel and that one, as most literary scholars do, he hopes to test hypotheses about the novel-in-general. Thus, whereas one of the goals of close reading, or “deep reading,” is the development of an internal, affective relationship to imagined others in the world, the goal of distant reading is an objective knowledge of world-scale systems, without the complications of the particular.
An exploration of an entity as enormous as the novel-in-general — at least without recourse to sweeping generalizations based upon a small and often unrepresentative cluster of texts — has only relatively recently become possible, thanks to large-scale digitization and markup projects and the development of advanced data mining tools and techniques. Such archives and methods and technologies permit scholars today to ask different kinds of questions about the materials they investigate. Moretti describes the approach to literary analysis that he develops across his essays as a “quantitative formalism”; like all formalisms, the point is not just the objective information derived — in this case, through counting — but rather what that counting reveals. As he argues in an exploration of the changing lengths and types of titles of novels between 1740 and 1850: “That titles became short is interesting, yes, but in the end, so what?” What Moretti does find interesting is “[t]hat by becoming short they adopted a signifying strategy that made readers look for a unity in the narrative structure.” The power of the quantitative thus lies in the new kinds of access to interpretation it provides.
Moretti acknowledges that, for him, the pinnacle of literary interpretation is formal analysis. For that reason, "what any new approach — quantitative, digital, evolutionary, whatever — must prove itself against" is precisely its ability to "do formal analysis, better than we already do." Better how, though? Better because broader; better because deeper; better because more comprehensibly mapped and graphed. But not necessarily more nuanced, and not without requiring, as Moretti’s quantitative formalism does, at times astonishing generalizations. Here is the crux, with which I do not think Moretti would disagree: the methods he develops are best suited to answering certain kinds of questions rather than others. Distant Reading inevitably raises the question not of whether one ought to read distantly, but of what one can read only distantly, and what one requires closeness in order to capture.
The discomfort of those who value closeness and depth in reading with the techniques and values of distant reading is, of course, not identical to the current anger in our culture with the revelation that our personal communications data is being mined. But the difference is precisely not about the techniques, which are substantially similar. It's not how the reading is being done that makes massive domestic surveillance problematic: it’s what is being read, and why it’s being read that way. Moretti is engaged in data-mining the public products of our shared culture. When an identically quantitative, computational approach is brought to reading personal data, and when the reason for reading it is not a clearer understanding of world-historical systems but instead the tracking of individuals, the very same techniques can be used in the service of a reading that is all too close. Techniques are techniques, in other words, neither good nor evil in themselves; the content under analysis and the purpose to which that analysis is put are, in the end, everything.