:: :: University of Virginia Library

One of the Cornell projects now well under way is a concordance to the complete writings of William Blake, edited by David Erdman of the New York Public Library. This concordance is based upon the variorum edition by Sir Geoffrey Keynes, but it will incorporate hundreds of corrected or added readings derived from a fresh collation of all Blake texts carried out over the past year by Dr. Erdman and his world-wide team of devoted Blakeians. The concordance will be, in effect, a new edition of Blake, albeit in somewhat scrambled order. All

lines which differ in any significant way from their counterparts in Keynes will be "flagged" in the concordance index and printed up separately as an Appendix to the volume. Since we are establishing as well as indexing a text, it is reasonable to suppose that a study of the concordance in its unpaged form (where changes are still possible) will help us in making final editorial decisions. Conjectural readings, for example, which we have arranged to accommodate, may be materially strengthened, or weakened, by evidence that turns up elsewhere in the text.

This integration of editorial and indexing routines is, I think, an important development, especially in a form which it might take in connection with any edition in progress. If an editor can arrange to finish his collations and read proof on the text before proceeding to the rest of his task, he can be provided with a concordance made from his own text to assist him in composing textual and critical notes, and the introduction to his volume. (Any editor will understand how valuable this assistance might be). The concordance could then be published at about the same time as the text itself, perhaps even as a companion volume. There is no reason why this procedure could not become entirely conventional. When it does, the punching of text on IBM cards, fast and simple as it is, will probably become obsolete. It is already possible to feed text into the computer directly from the perforated tapes produced by an ordinary monotype machine; it may soon be possible to scan print photoelectrically and transfer it directly to magnetic tape for computer processing.

This is one of the directions that computer work will inevitably take within a few years. To illustrate another I shall again present a single example in the hope that wider inferences may be drawn from it. In the field of stylistic analysis a whole new world of possibilities seems about to open up, the shape of it already discernible. A unique feature of the Arnold concordance (and one which we expect to furnish on all the Cornell concordances) is a list of index words in order of frequency, produced for us by the computer. But counting frequencies is only one of many operations a computer might be expected to do by way of analyzing characteristics of a literary text. The computer's insensitivity to anything but physical characteristics is a smaller handicap than one would imagine, for it can still do something like the things we do ourselves when we identify style. Where we may observe that Samuel Johnson wrote in rotund oratorical sentences and used a Latinate vocabulary, the computer would measure the unusually long intervals between spaces and between periods, and would record the high

frequency of commas and of letter patterns like "MENT," "TION," "ATE," "PRE," and the like. Where we might notice, and mark as characteristic, certain conceptual relationships in Johnson's thinking, the computer, with its infallible memory, might accumulate definitive evidence of a kind to which we are normally insensitive, such as keyword clusters, syllable frequencies, trigraph patterns, verb-noun ratios, and other concrete properties which every concept takes on when it is given literary form. When the computer is dealing with an unknown text, the properties it measures might be matched against analogous properties of known texts, and a series of weighted scores assigned to show the degree of correspondence. Thus — to become wholly fanciful for a moment — a newspaper sonnet of unknown authorship might yield a score of, say, 35 as Coleridge, 39 as Wordsworth, 28 as William Lisle Bowles, or Charles Lloyd, or Mrs. Mary Robinson, but a score of, say, 71 as Southey. On grounds like these, provided that there is no external evidence that is contradictory, we would be tempted to attribute the sonnet to Southey.

Now the critical imagination may shudder at the thought of running enough tests on the sonnets of Bowles, or Charles Lloyd, or Mrs. Robinson to build up the necessary bank of scores. But there is likely to be a far more serious objection to the procedure I have fanci-fully outlined — even if the results could be made to turn out cleanly. Some readers will perhaps recognize that what I have described resembles a cryptanalytic attack on a piece of cipher text, and will be properly skeptical of its validity as applied to literary text. For as Colonel and Mrs. William Friedman have recently reminded us, in their brilliant and amusing account of the search for cipher in Shakespeare,[6] a cryptanalytic attack is valid only if an underlying system does in fact exist in the text. And who will declare that a system exists in a man's literary style? When we measure the properties of style we measure the man himself, his reason, his logos. Surely what I have proposed is a more terrible thing even than any possible menace to our human supremacy exerted by electronic brains!

Yet I am prepared, I think, to press the idea. When one considers soberly the progress we have made during the last half-century in measuring attributes of the mind, we would be incautious indeed to conclude that we have reached the end of this investigation. More probably, we have only begun. But to put the matter in these terms at all is less realistic, I suggest, than to regard the computer technique as simply an extension of the kind of stylistic analysis now being practiced.

If the computer does most of the things we do ourselves when we analyze style, the crucial difference is that the powers of this sort of analysis can be fantastically multiplied by electronic application, even with machines now available. And we must understand that we are just entering the age of computer technology. Today, hardly more than 15 years from the opening of this age, computers regularly become obsolete as rapidly as they can be built. I have already mentioned input devices that are learning to read photoelectrically; operating speeds are rising astronomically as tubes give way to transistors, and transistors to lowtemperature crystals (the new IBM 7074 is twenty times as fast as the 7070, itself a transistorized machine and some three times as fast as our lumbering old 704); computer memories are expanding to accommodate hundreds of thousands of words; programs are becoming sophisticated enough to perform accurate and grammatical translation. All these developments, and others equally breath-taking, suggest that it cannot be long before computers will undertake successfully the most delicate and complex programs of stylistic analysis.

Nor can it be long, I trust, before the research scholar in the Humanities will recognize these developments and learn to turn them to his advantage — to venture more freely across the boundary that, lamentably, separates his culture from that of the scientist. I am not suggesting that we should celebrate the coming of a new god (ex machina, naturally) as Yeats might have done, had he lived into the age of cybernetics and "information theory" — perhaps seeing the electronic brain as a great smooth beast, with "gaze blank and pitiless," rolling evenly towards New York to be born. I am only suggesting that the "scientific revolution," which is being created without much help from us, is probably the greatest single fact of our century; that it will go on expanding whether we recognize it or not; that we have nothing to fear and everything to gain from coming to terms with it; and that if we learn to exploit its potentialities we shall be serving the cause of the Humanities in the best possible way.