Of Ngrammatology

I’ve recently discovered Google’s Ngram Viewer. If you haven’t found and played with it yet, you will.

The Ngram Viewer takes a corpus of just over five million library books digitized by Google and, within that arena, instantly searches for terms or phrases you may want to explore, tabulating the frequency of their occurrences over time.

The Ngram algorithm might let you visualize, for example, 20th-century deployments  of the words nitpicker, caviller, and momus—to choose more or less at random three epithets that might be applied ungenerously to a writer on language. You may already use nitpicker in this sense. (If you have grade-school children you may have another—more visceral—sense of the word, as well.)

Caviller is a pretty unusual word in contemporary discourse. But if you know the verb to cavil, meaning to insist on making pointlessly small distinctions, you can infer that caviller is one who cavils.

Momus—the mythological spirit of derisive contempt—is rarer still. No one wishes to be described as a momus.

An Ngram snapshot of these three sorry terms shows that momus occurs infrequently in Google’s library books, while caviller was doing pretty well around 1900 before going into steep decline thereafter.

Nitpicker gets going about the time I did, in the middle of the last century. If the Ngram is to be trusted, nitpicker is steaming ahead. I find this strangely comforting.

The Ngram Viewer lets you plug in phrases, too. You can enter and chart the fortunes of Queen Victoria’s most famous sound bite, provided that you realize a) Victoria Regina may never have said any such thing, and b) the algorithm will select for any occurrence of that string, not only those tied to the Widow of Windsor.

“We are not amused” seems to have peaked in the 1930s. The 30s may have been the 20th century’s least amusing decade, though it has some pretty stiff competition.

Ngrams are also good for things besides terms or phrases. For some time, scholarly curiosity has been whetted by statistical analyses demonstrating the frequency with which the names of influential thinkers are invoked, if not spiritually then at least bibliographically.

PMLA, the journal of the Modern Language Association, used to make something of citation stats, which I always found rather depressing, since a small number of critics would show up over and over (and I say this even though in my previous line of work I published a number of them). Alas, the Ngramminess of Freud, Marx, and Foucault—to name three of criticism’s Eumenides—map a tale of downward trending.

Somewhere in library heaven, Marx is wistful for 1975, Freud for 1992, Foucault for 1996. Them were the days.

Them were, of course, theory’s days. But theory’s constellation is now low on our critical horizon. The ascendant figures are cognitivism and Big Data, for which the Ngram Viewer is an irresistible tool.

Statistical analyses, from word clusters to distant reading, are now the surveillance satellites and bathyscaphs of the practice we used to call literary analysis. We have become charters and mappers, linguistic cartographers in search of the texts we thought we knew how to find and how to read.

On one level, there’s no reason to mourn theory, since like the figures in the zodiac’s cast of characters, it was never quite alive to begin with. But as we take up the newest tools of analysis—including the seductions of the Ngrammatological—it’s useful to remember that these constellations of ideas are always moving across our heavens, to sink and to return.

Ngramminess is just one more feature of our data-verse. (Somebody really should give awards out for the best statistical analysis of literature using the Ngram Viewer. They  would, of course, have to be called the NGrammies.)

What our remorseless turn toward counting and mapping will tell us about literature, however, only time will tell. And what it will tell, the humanist in me continues to think, will ever be open to interpretation.

@WmGermano