David Hume said that the association of ideas depended on contiguity, resemblance, or causality. Of these possible connections contiguity --or metonymy in the parlance of literary criciticsm --is much the dumbest. The stupidity of mere juxtaposition is indeed a scandal to the mind, whose program could be summarized as: Where metonymy was, there metaphor and causality shall be. "Post hoc ergo propter hoc" is an old caution against the excesses of this desire. When people are too successful in the transformation of metonymies we call them paranoid, and I am certainly not the first person to call paranoia the professional disease of highly intelligent people.
If metonymy is such a scandal why has so much energy gone into a reverse engineering that turns causality and metaphor into metonymy? Wherever a dictionary or index is made, a complex order of some kind is turned into an alphanumerical order of a much simpler and more arbitrary kind. Goethe's devil, speaking at the end of the great taxonomic century to which we owe Linnaeaus, the Encyclopédie, and Johnson's Dictionary, was very clear on this point:
| Wer will was Lebendigs erkennen und beschreiben Sucht erst den Geist herauszutreiben, Dann hat er die Teile in seiner Hand, Fehlt, leider! nur das geistige Band. |
Who would study and describe the living,
starts By driving the spirit out of the parts: In the palm of his hand he holds all the sections, Lacks nothing except the spirit's connections |
Faust, Part One, 1936-39
The original sinners in this enterprise were the medieval monks who transformed the Living Word of God into the alphabetically contiguous places known as a "concordance." Their reduction of the Bible into an inventory of its parts was inspired by the desire to create finding tools for imperfect human minds. Thus the decomposition of a spiritual order of words into mere alphabetical contiguity is also a spur to further thought. There is no room for paranoia in interpreting the Word of God since everything is connected to everything. If the Bible uses a word twice, there must be a divine reason for it. If my memory is too weak to remember one passage when I read the other, the metonymic reduction of the concordance corrects my failure and directs my attention to two dots that it is my task to connect in a spirit of charity. The concordance is named after the goal that inspires its creation: it is a tool for discovering the concord of the universe of the Word.
Something like this may serve as the hermeneutical horizon for looking at digital philology. Computers emphatically do not think. They are the metonymic machines par excellence and do not understand the first thing about resemblance or causality unless it has first been resolved into minute steps of contiguity. But their immense stupidity allows an intelligent reader to decompose complex textual structures into an indefinite number of arbitrary metonymic structures that can act as a spur to thought.
I have used the example of the Bible partly because it is historically appropriate and partly because my project is focused on Shakespeare, the most Scriptural of secular authors. I do not want to claim that every word in Shakespeare is connected to every other word. It is enough to claim that many words are connected to many words and that many of those connections in turn become visible through a digitally assisted decomposition of the text into new metonymies.
A less fancy way of putting this is to say that with a properly tagged electronic text you can explore the verbal neighborhoods of Shakespeare's plays from a variety of perspectives. The filter and sort routines of business databases provide the techniques for such explorations. They are particularly helpful in letting you look for unknown entities that meet certain parameters, such as a list of words that only occur in Hamlet. The adaptation of business technology to literary purposes, incidentally, is an old story: the Greeks got their alphabet from the Phoenicians, and it was almost certainly used by traders before it was used by poets.
The text as database can be used to generate lists in which the order of words is not that of the text. To read such lists is a way of reading the text against the grain in a deliberately "mindless" fashion. Ancient readers opened sacred texts like the Bible or the Aeneid at random to seek guidance --a perfectly sensible procedure in a world where "there is special providence in the fall of a sparrow." In such a world a list of ten words drawn at random from a database of the Bible would present me with a challenge to understand the meaning of their revelation to me at this point and in their particular order.
If I live in a probabilistic universe I am most unlikely to look for revelations in such a list. On the other hand, a proper random sample of words from the Bible generates a list from which some very general propositions about the text are more easily apparent than from any text string of comparable length. While random lists have their uses, biased lists are much better for special purposes. I have a particular interest in the ways in which Shakespeare's plays neighbor each other, and I want to document the scope of these neighboring relationships from a quantitative and qualitative perspective. For this purpose it is useful to generate lists in which contiguity is the cause of selection. An example would be a list of words that occur in Hamlet, in Measure for Measure, and in no more than two other works by Shakespeare.
Unless you believe that Shakespeare's verbal universe is ruled by special providence there is no reason to assume that the pure metonymy of this list must contain an interpretable meaning. But the list becomes a useful tool on the much more modest assumption that the quantity and quality of rare shared vocabulary is a promising measure of the mutual proximity of Shakespearean plays. You may learn something from the relative lengths of such lists. More interestingly, any such list holds out the hope that its dumb metonymies are a quarry that can be successfully mined. For while from one perspective the list is entirely mindless, from another perspective it was designed with the purpose of reducing randomness: if there is anything to the hypothesis that close neighbors share more rare vocabulary, then a list of such vocabulary should hold a high concentration of interpretable results. And in fact this seems to be the case sufficiently often to make it a worthwhile activity to read against the grain in this fashion, survey the verbal neighborhoods by scanning lists of this kind, and transform their metonymies into resemblance wherever the evidence is plausible enough to support such a claim.
The Hamlet and Measure for list, for instance, contains among its thirty odd items the morphologically related cluster of 'overdone', 'overweigh', 'unpregnant', and 'unshape'. For a variety of reasons, this is not a cluster easily picked up by reading the plays, but it jumps out from a reading of the list, and lexical overlap in this case leads to thinking about interesting overlaps between the roles of Hamlet and Angelo. The utterly unremarkable word 'pop', together with 'legitimate' and 'Nero' weaves a web that strongly ties the Falconbridge plot of King John to themes of adultery and disinheritance in Hamlet and Lear. The list of such clusters is not endless, and it would be foolish to claim that they hold they key to previously undisclosed mysteries. If they did, the method would convict Shakespeare (or any poet to whom it is applied) of incompetence. On the other hand, the mapping of verbal neighborhoods is a pretty good way of demonstrating how deeply and extensively the thematic structures of the plays are woven into its linguistic texture at the most mundane and microscopic levels.
19 October 1999