The Bactra Review: Occasional and eclectic book reviews by Cosma Shalizi   64

Atlas of the European Novel


by Franco Moretti

London and New York: Verso, 1998.

One Effort More, Litterateurs, if You Would Be Empiricists!

Franco Moretti is a rare bird, perhaps unique: a full-fledged advanced, "theoretical" literary critic; a clear and enjoyable writer, witty when he wants to be; someone with good taste in books (i.e., it often agrees with mine) and generally sensible ideas (he was nearly the only one of his kind to take the correct side in the Sokal affair); and a preserver of the old left-wing fondness for science, in a modest way (he quotes Stephen Jay Gould's popular essays and The Structure of Scientific Revolutions intelligently, has at least leafed through Popper, and keeps mum about quantum mechanics, relativity, and Newton). This fondness for science (I will not say scientism) shows itself in his concern with the evolution, in a pretty strict sense, of literary forms. It also shows up in his penchant for quantifying, one of the more piquant instances of which is a footnote in The Modern Epic which shows that the distance of countries from Ireland is positively correlated with the frequency with which references to them in Ulysses are cliches. (I can't remember whether Moretti thinks this is a reflection on Bloom or on Joyce.) The former topic, while extremely interesting, will have to wait for another time, when my copies of the relevant books are not rusticating in Maryland; in his new Atlas, on the other hand, he has abandoned himself to the later propensity, and the book is before me as I type.

The first two chapters consider geography in the novel: where characters in selected novels are from, and where the action happens. This is normally left to fans, producing things like the atlas of Patrick O'Brian's Aubrey-Maturin novels. Moretti wants to milk more out of this than just following particularly tricky engagements of the H.M.S. Surprise, however: he ties the rise of the novel to that of the nation-state in the first chapter, class structure to urban novels in the second. (Jacques Barzun somewhere complains that the middle class is like a helium balloon, always rising: so are the novel and the nation-state.) There are many curious and interesting facts in these chapters; also an argument with Edward Said (his departmental chairman) over Mansfield Park, which I think Moretti wins. Though diverting, these chapters don't really reveal that much. At the end of the day, Moretti is saying that different classes of people live in different places, and that the novels he looked at mostly reflected this accurately; with the exception of villains in English novels, who tend to be foreigners, especially Frogs. (Did French novels of this period import their villains from perfidious Albion? What were the European trade-balances in villains? Was there specialization and division of labor, as economic theory leads us to hope --- this country producing seducers of the innocent, that one criminal masterminds, a third thuggish henchmen? Inquiring minds, etc.) One may be pardoned for meeting these revelations with a nod. That the action in historical novels concentrates at borders and difficult, untamed pieces of terrain is more interesting. Moretti says this has something to with "the need to represent the territorial divisions of Europe"; some readers will mutter "whose need?", "wasn't that what they had maps for?" and "plausible places for adventures, ain't they?"

The really interesting stuff is in the third and last chapter, when Moretti gets into the geography of novels, the distribution of the bound, ink-splattered hunks of wood-pulp in space and time. He starts by looking at English circulating libraries (including, without comment, a colonial one in Madras), and makes an interesting observation: the smaller the library, the more its repertoire of novels conforms to the canon. Moretti interprets this as market forces at work, urging small firms (and the circulating libraries were for-profit institutions) to play safe: "If there is only one book, Religion. And if there is only one bookcase, the Canon." More: for the English libraries, the English canon; the percentage of foreign novels out of the total is small in all cases, negligible for some of the smaller libraries. This leads to an examination of translations in different national bibliographies: these are least, in proportion to the total, for Britain and France, and in the case of Britain show a steady decline, mirrored in the records of the circulating libraries. (The French record is more even, at least until about 1840.) Moretti concludes that what was read was fairly homogeneous across England; was mostly English; and, as was generally the case in all European countries, was mostly produced in the metropolis. (That wealth and power attract talent was old news in Uruk when Gilgamesh was king.)

What, then, about the English novel abroad? Moretti takes samples of five British authors and ten different genres, and looks to see how much of each sample was translated into different national literatures --- French, Dutch, Danish, Spanish, Italian, Polish, Hungarian, Rumanian; and again for six French authors (lumping Stendhal and Balzac together), French sentimental novels, and "minor historical novels." The biggies --- Scott, Dumas, Bulwer-Lytton (he of the bad-opening-sentence contest), Dickens, Hugo, Sue --- these go everywhere; British "sensation novels" and French sentimental novels do about as well. (Exception: Scott bombs with the Magyars, "one or two sporadic translations." Why?) Below this scale of pan-European approval, what's translated varies a lot: the Dutch get all the English sea-stories and most of the oriental tales, the Rumanians none of either. This variation has, says Moretti, two causes. First, these countries have smaller literary markets, and just can't take everything the English and French produce, plus the output of their own writers; some works will be dropped by happenstance. (How do we know that Italy had a smaller literary market than England?) Then, too, there are active tastes at work, cultural affinities that give the Italians English religious novels and keep Stendhal and Balzac out of Britain. In any case, there's a definite center of production, Britain and France; a second tier of nearby countries which absorb most of what they produce; and a third tier of more disconnected countries. This resembles the pattern of development and integration into the world markets (intentionally: Moretti is a fan of Immanuel Wallerstein's "world-system" notions), and the tiers in fact correspond, roughly, to the degree of industrialization. Cultural diffusion isn't new, as he admits; but he wants to claim that diffusion from such a marked center is. (That is most disputable, certainly for products of literate cultures.) He closes by reflecting on how new literary forms have a way of appearing at the periphery, instead of the productive center --- the novel of ideas (Russia); magical realism (Latin America, slighting Ray Bradbury); even the historical novel, in Scotland.

Non-fiction, for me, can be split neatly down the middle, into works of reflection and works of conviction. A good work of reflection appeals to the imagination: it seizes and holds its reader's attention, its style pleases, it provokes contemplation, "the play of musement," flights of fancy; it provides what we misleadingly call insights. Such works are an integral part of any real literary culture, and the ability to enjoy them is one of the chief benefits conferred by a liberal education. Works of conviction, on the other hand, aim at just that, securing the belief of their readers, and the virtues proper to them are sound facts, strong arguments, valid reasoning, shrewd inferences, inescapable conclusions, measured doubts. If it is well-written, so much the better.

Much as I like works of reflection, I can't take them seriously. I read them for no loftier reason than pleasure (which can of course be had from tragedy and jeremiad as well as from less somber confections). I can be caught up in what they say; I don't give it credence. It is in this spirit that I read almost all literary studies, enjoying the ride and ignoring the pretensions to establishing various hypotheses. (I used to feel bad about this, but then a literary critic I know told me that he does much the same.) Judged as such, Moretti's book is very fine, with plenty for anyone with a modicum of interest in European literature, and another modicum of analytical curiosity, to turn over in their mind.

Moretti would not be content with this. He wants to produce a successful work of conviction. I don't think he has; in particular, he falls far short of acceptable standards of statistical hygiene and inferential probity.

Let's start an aspect of "tact in drawing inferences" crucial to comparative studies, asking, with Lichtenberg, "You have discovered these traits together ten times, but have you counted the times you have not found them together?" (More pungently: "We must not seek to abstract from the busts of the great Greeks and Romans rules for the visible form of genius as long as we cannot contrast them with Greek blockheads.") Moretti is much, much better about this than almost all of his colleagues; but even he stumbles far too often. "Near the border, figurality goes up. Beyond the border, it subsides" (p. 45): such is the burden of the first "theoretical interlude," on "space and style" (pp. 40--7). Its basis? Two contrasting passages from Waverley, and another pair from Notre Dame de Paris. (Moretti isn't bothered that there is no border invoked in Hugo; that the contrast in figurative language between Hugo's two passages is slight; that he himself equivocates between figures of speech in general and metaphors in particular; neither will we be bothered by these little lapses.) But this of course doesn't even establish the correlation for those two novels, let alone any class of novels to which we take them both to belong (e.g., ones whose authors' surnames include the letter O). How often does Scott indulge in figurative language? How often can Hugo restrain himself from it? How often does either story approach borders? How many figures of speech should we expect on those occasions, if the former are evenly distributed through the text? Answering these questions would involve a lot of boring work, starting with flagging all the metaphors in the book. That's life, or anyhow scholarship; but how much more pleasant to vaguely link it to nationalism, to Ricoeur's (rather unoriginal) ruminations on on the cognitive role of metaphors, to liminal things of all kinds.

Moretti wants to find the frequencies with which different kinds of book occur in large libraries, by means of small samples. This is a perfectly legitimate procedure, but it is of course very unlikely that the frequencies found in a sample of, say, 100 books are exactly the same as the frequencies in the entire population of, say, 1,000 books. To make comparisons between different populations, we need to know how precise the estimates are: the difference between 60% and 50% doesn't signify if the uncertainty in each case is plus or minus 10%. What Moretti did was "sampling without replacement," and calculating the confidence intervals for this procedure is an old chestnut of statistical inference. (For fixed sample size, unsurprisingly, the confidence intervals are tighter for small populations than for large ones. Conversely, and again unsurprisingly, for fixed populations, smaller samples have larger confidence intervals.) It'd be unreasonable to expect Moretti, a middle-aged full professor, to learn statistical inference, no matter how much he writes in praise of "serial history." Whether it is equally unreasonable to expect him to send a graduate student down to the college bookstore with a twenty for Schaum's Outline of Statistics and a roborative latte, look up "Proportions, confidence intervals for" in the index, turn to p. 196, and start plugging and chugging, is another matter.

For example: We take a sample of 100 from a population of 2500, and find that 50 members of our sample have some property C. With 95% confidence, we can say that the fraction of the population which is C is between 0.40 and 0.60. From a different population of 1000, we take a sample of 100 and get 65 C's, and know, with the same confidence, that in that population the frequency is between 0.56 and 0.74. Is the difference between populations significant? Yes, but at just under the 27% level, i.e., there's just under a 27% chance of getting such a discrepancy by sheer chance; and even the social sciences typically won't touch anything above 5%.

For C, read percentage of "novels ... written by authors listed in the Dictionary of National Biography" (p. 145--6), Moretti's proxy for the literary canon; lacking precise frequencies and population sizes, I guessed them from his fig. 69 (p. 147). All told, the calculations took five minutes on a pocket calculator, not counting going to the library for the table of significance levels in the back of the aforementioned Schaum's.

Let us suppose --- what seems likely --- that Moretti's negative correlation between library size and canonicity survives, after we take account of the uncertainties in his estimates, and is significant at some modest but real level, say, 10%. Even then we are not done. Random samples from a population typically contain less variation than the population itself, and this selection is stronger the smaller the sample is. (This is easy to see in the extreme case where the sample consists of a single item.) Look at his maps of the distribution of translations of different genres of English and French novels. Now, two questions. What fraction of the total of commercially successful novels do these genres represent? (Successful in their original language, I mean; no other novels were or are likely to be translated.) How much of the reported variation can be accounted for by the distortions of sampling? Moretti doesn't answer the first question, and the second doesn't enter his mind.

Moretti has spoken in several places of his desire for a "falsifiable" literary criticism, by which he means one which can hold up its head in empirically respectable company. (As is typical of his profession, he writes as though the philosophy of science ended in 1970.) He's betting that quantification is one way of pursuing this laudable goal. Well and good; "a mighty fortress is our mathematics," as Ulam said, and it can handle just about anything, but raw quantification is not enough. If numbers were knowledge, then IQ-mongers would not be idiots. To weld mere numbers into reliable conclusions requires tools; in this case, those of statistics.

Ignorance of statistics, even to the point of confusing a mean with a median, is as inconsequential as not knowing Islamic jurisprudence, or never having read Tolstoy. Happy, fulfilled and useful lives have been lead by unnumbered millions who were similarly unaccomplished, who'd not have been the least bit better off if those deficiencies had been remedied. Things change, of course, if one goes to court in Riyadh, teaches a seminar on Anna Karenina, or pursues an essentially statistical line of inquiry. Moretti has embarked on the last-named, and while I'm almost sure he can tell a mean from a median (almost, because it doesn't come up in his Atlas), he is, by his own confession, a statistical tyro, and, more seriously, seems to have taken no steps to remedy this.

No one would blame him if those steps meant arduous and disfiguring studies, best left to the young and perverse. It would be unreasonable to expect a literary critic, in mid-career, to master the theory of formal languages, automata, and abstract computation: it would be a lot of work, and why should he bother? It's not like that theory has ever told us anything interesting about literature. (He can, however, refrain from saying that assigning characters to partially-overlapping classes constitutes a "generative model" of a novel, even, or rather especially, on Bourdieu's say-so.) But in fact the statistics he needs aren't very abstruse; countless fools have learnt them; "what one fool can learn to do, another can"; and Moretti is no fool.

Having thus tried to teach one of my betters his business, I'll finish by being even more patronizing: Moretti shows great promise. His next book may even carry conviction.

xiv + 206 pp., 91 black and white figures, index of names and titles (but not, most vexingly, subjects), bibliographic footnotes.

Europe / Fiction / Geography / Literature and Literary Criticism / Modern History

Currently in print as a a hardback, US$22, ISBN 1-85984-883-4 [buy from Powell's], and as a paperback, US$, ISBN 1-85984-224-0 [buy from Powell's], LoC PN3383.S67 M67. First published as Atlante del romanzo europeo 1800--1900, Turin: Einaudi, 1997; no translator is listed.

21 October 1998